Insilico Medicine
LongevityBench: Aging Biology Across Omics Domains
Seventeen aging-biology tasks across clinical, epigenomic, transcriptomic, proteomic and genetic data.
aging-biologymultiomicstexttabular
Status
curated
Saturation
unknown
Reproduction
not yet
Published results
442 results
Editor's note
Official results cover the site's 26 visible models, including fine-tuned Longevity-LLMs and untuned base/frontier models. Per-task metrics retain native units and direction; scores are not averaged into a BioInspect index. Model training exposure and task-specific scoring should be considered when comparing specialized models with general-purpose models.
Published results
Results are separated by evaluation version and source configuration. Scores in different sets are not a single comparable ranking.
Official per-task values, with source-reported SE in the same units as the score. MAE is in years (lower is better); balanced accuracy and concordance are proportions (higher is better). Source prompt counts are not inferred valid-response denominators. No answer-format validity or solved counts are inferred. The site's aggregate ranks use optimistic valid-only scoring and equal-domain weighting; no aggregate rank is calculated here.
NHANES Age / Multiclass
| Model | Score | Variant | Source |
|---|---|---|---|
| Claude-Opus-4.6 | 0.373 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Claude-Opus-4.6 |
| Gemini-3.1-Pro | 0.365 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Gemini-3.1-Pro |
| Gemini-2.5-Pro | 0.355 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Gemini-2.5-Pro |
| Claude-Opus-4.5 | 0.353 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Claude-Opus-4.5 |
| L-Qwen3-1.7B | 0.346 balanced accuracy ± 0.007 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / L-Qwen3-1.7B |
| L-LFM2-2.6B | 0.345 balanced accuracy ± 0.007 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / L-LFM2-2.6B |
| GPT-5 | 0.344 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / GPT-5 |
| GPT-5.2 | 0.337 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / GPT-5.2 |
| Grok-4.3 | 0.330 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Grok-4.3 |
| Claude-Sonnet-4.6 | 0.330 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Claude-Sonnet-4.6 |
| Kimi-K2.5 | 0.329 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Kimi-K2.5 |
| Gemini-2.5-Flash | 0.328 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Gemini-2.5-Flash |
| GPT-5-mini | 0.310 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / GPT-5-mini |
| L-Qwen3.5-9B | 0.305 balanced accuracy ± 0.007 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / L-Qwen3.5-9B |
| L-Qwen3-0.6B | 0.305 balanced accuracy ± 0.007 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / L-Qwen3-0.6B |
| DeepSeek-R1 | 0.301 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / DeepSeek-R1 |
| Claude-Sonnet-4.5 | 0.295 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Claude-Sonnet-4.5 |
| L-LFM2-1.2B | 0.289 balanced accuracy ± 0.007 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / L-LFM2-1.2B |
| GPT-4.1 | 0.286 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / GPT-4.1 |
| GPT-4.1-mini | 0.277 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / GPT-4.1-mini |
| DeepSeek-V3.2 | 0.263 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / DeepSeek-V3.2 |
| Qwen3-1.7B (base) | 0.230 balanced accuracy ± 0.006 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Qwen3-1.7B (base) |
| LFM2-2.6B (base) | 0.214 balanced accuracy ± 0.006 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / LFM2-2.6B (base) |
| Qwen3-0.6B (base) | 0.179 balanced accuracy ± 0.006 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Qwen3-0.6B (base) |
| Qwen3.5-9B (base) | 0.178 balanced accuracy ± 0.006 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / Qwen3.5-9B (base) |
| LFM2-1.2B (base) | 0.155 balanced accuracy ± 0.006 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_multiclass / LFM2-1.2B (base) |
NHANES Age / Pairwise
| Model | Score | Variant | Source |
|---|---|---|---|
| L-LFM2-2.6B | 0.807 concordance ± 0.009 SE | L-LLM · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / L-LFM2-2.6B |
| L-Qwen3.5-9B | 0.803 concordance ± 0.009 SE | L-LLM · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / L-Qwen3.5-9B |
| Gemini-3.1-Pro | 0.797 concordance ± 0.009 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Gemini-3.1-Pro |
| L-LFM2-1.2B | 0.787 concordance ± 0.009 SE | L-LLM · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / L-LFM2-1.2B |
| Claude-Opus-4.6 | 0.782 concordance ± 0.009 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Claude-Opus-4.6 |
| L-Qwen3-0.6B | 0.774 concordance ± 0.009 SE | L-LLM · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / L-Qwen3-0.6B |
| Claude-Opus-4.5 | 0.769 concordance ± 0.009 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Claude-Opus-4.5 |
| GPT-5.2 | 0.752 concordance ± 0.009 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / GPT-5.2 |
| Gemini-2.5-Pro | 0.741 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Gemini-2.5-Pro |
| Kimi-K2.5 | 0.737 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Kimi-K2.5 |
| L-Qwen3-1.7B | 0.734 concordance ± 0.010 SE | L-LLM · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / L-Qwen3-1.7B |
| Grok-4.3 | 0.730 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Grok-4.3 |
| Claude-Sonnet-4.5 | 0.724 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Claude-Sonnet-4.5 |
| Claude-Sonnet-4.6 | 0.716 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Claude-Sonnet-4.6 |
| GPT-4.1 | 0.713 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / GPT-4.1 |
| GPT-5 | 0.705 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / GPT-5 |
| GPT-5-mini | 0.693 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / GPT-5-mini |
| Gemini-2.5-Flash | 0.690 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Gemini-2.5-Flash |
| DeepSeek-R1 | 0.679 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / DeepSeek-R1 |
| GPT-4.1-mini | 0.650 concordance ± 0.010 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / GPT-4.1-mini |
| LFM2-2.6B (base) | 0.570 concordance ± 0.011 SE | base · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / LFM2-2.6B (base) |
| Qwen3.5-9B (base) | 0.551 concordance ± 0.011 SE | base · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Qwen3.5-9B (base) |
| DeepSeek-V3.2 | 0.537 concordance ± 0.011 SE | frontier · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / DeepSeek-V3.2 |
| LFM2-1.2B (base) | 0.522 concordance ± 0.011 SE | base · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / LFM2-1.2B (base) |
| Qwen3-0.6B (base) | 0.511 concordance ± 0.011 SE | base · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Qwen3-0.6B (base) |
| Qwen3-1.7B (base) | 0.489 concordance ± 0.011 SE | base · 2102 source prompts | Insilico Medicine · LongevityBench nhanes_age_pairwise / Qwen3-1.7B (base) |
NHANES Age / Regression · lower is better
| Model | Score | Variant | Source |
|---|---|---|---|
| Gemini-3.1-Pro | 12.1 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Gemini-3.1-Pro |
| GPT-5.2 | 12.2 yr mean absolute error ± 0.1 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / GPT-5.2 |
| Kimi-K2.5 | 12.4 yr mean absolute error ± 0.1 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Kimi-K2.5 |
| Claude-Opus-4.5 | 12.5 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Claude-Opus-4.5 |
| Claude-Opus-4.6 | 12.7 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Claude-Opus-4.6 |
| Gemini-2.5-Pro | 12.8 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Gemini-2.5-Pro |
| Grok-4.3 | 13.0 yr mean absolute error ± 0.1 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Grok-4.3 |
| Claude-Sonnet-4.6 | 13.1 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Claude-Sonnet-4.6 |
| GPT-5 | 13.3 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / GPT-5 |
| GPT-5-mini | 13.5 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / GPT-5-mini |
| DeepSeek-V3.2 | 13.9 yr mean absolute error ± 0.1 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / DeepSeek-V3.2 |
| L-Qwen3-0.6B | 14.0 yr mean absolute error ± 0.2 yr SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / L-Qwen3-0.6B |
| Gemini-2.5-Flash | 14.0 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Gemini-2.5-Flash |
| GPT-4.1 | 14.2 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / GPT-4.1 |
| Claude-Sonnet-4.5 | 14.3 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Claude-Sonnet-4.5 |
| L-LFM2-2.6B | 14.8 yr mean absolute error ± 0.2 yr SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / L-LFM2-2.6B |
| GPT-4.1-mini | 15.1 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / GPT-4.1-mini |
| L-Qwen3-1.7B | 15.7 yr mean absolute error ± 0.2 yr SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / L-Qwen3-1.7B |
| DeepSeek-R1 | 15.9 yr mean absolute error ± 0.2 yr SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / DeepSeek-R1 |
| Qwen3-1.7B (base) | 16.4 yr mean absolute error ± 0.2 yr SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Qwen3-1.7B (base) |
| Qwen3.5-9B (base) | 16.5 yr mean absolute error ± 0.2 yr SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Qwen3.5-9B (base) |
| L-Qwen3.5-9B | 17.0 yr mean absolute error ± 0.2 yr SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / L-Qwen3.5-9B |
| LFM2-2.6B (base) | 20.8 yr mean absolute error ± 0.2 yr SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / LFM2-2.6B (base) |
| L-LFM2-1.2B | 21.1 yr mean absolute error ± 0.2 yr SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / L-LFM2-1.2B |
| LFM2-1.2B (base) | 21.2 yr mean absolute error ± 0.2 yr SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / LFM2-1.2B (base) |
| Qwen3-0.6B (base) | 24.0 yr mean absolute error ± 0.3 yr SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_age_regression / Qwen3-0.6B (base) |
NHANES Mortality / Binary
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3-0.6B | 0.890 balanced accuracy ± 0.005 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / L-Qwen3-0.6B |
| L-LFM2-2.6B | 0.887 balanced accuracy ± 0.005 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / L-LFM2-2.6B |
| L-Qwen3.5-9B | 0.885 balanced accuracy ± 0.005 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / L-Qwen3.5-9B |
| L-LFM2-1.2B | 0.885 balanced accuracy ± 0.005 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / L-LFM2-1.2B |
| Gemini-3.1-Pro | 0.883 balanced accuracy ± 0.005 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Gemini-3.1-Pro |
| L-Qwen3-1.7B | 0.881 balanced accuracy ± 0.005 SE | L-LLM · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / L-Qwen3-1.7B |
| GPT-5 | 0.874 balanced accuracy ± 0.005 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / GPT-5 |
| GPT-5-mini | 0.873 balanced accuracy ± 0.005 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / GPT-5-mini |
| LFM2-1.2B (base) | 0.848 balanced accuracy ± 0.006 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / LFM2-1.2B (base) |
| GPT-5.2 | 0.844 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / GPT-5.2 |
| Claude-Opus-4.6 | 0.834 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Claude-Opus-4.6 |
| GPT-4.1 | 0.831 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / GPT-4.1 |
| Claude-Sonnet-4.5 | 0.827 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Claude-Sonnet-4.5 |
| Claude-Opus-4.5 | 0.817 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Claude-Opus-4.5 |
| Claude-Sonnet-4.6 | 0.811 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Claude-Sonnet-4.6 |
| Kimi-K2.5 | 0.800 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Kimi-K2.5 |
| GPT-4.1-mini | 0.785 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / GPT-4.1-mini |
| Grok-4.3 | 0.773 balanced accuracy ± 0.006 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Grok-4.3 |
| Qwen3-0.6B (base) | 0.705 balanced accuracy ± 0.007 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Qwen3-0.6B (base) |
| DeepSeek-V3.2 | 0.700 balanced accuracy ± 0.007 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / DeepSeek-V3.2 |
| DeepSeek-R1 | 0.556 balanced accuracy ± 0.008 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / DeepSeek-R1 |
| Gemini-2.5-Pro | 0.539 balanced accuracy ± 0.008 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Gemini-2.5-Pro |
| Qwen3-1.7B (base) | 0.516 balanced accuracy ± 0.008 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Qwen3-1.7B (base) |
| Gemini-2.5-Flash | 0.448 balanced accuracy ± 0.008 SE | frontier · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Gemini-2.5-Flash |
| Qwen3.5-9B (base) | 0.424 balanced accuracy ± 0.008 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / Qwen3.5-9B (base) |
| LFM2-2.6B (base) | 0.374 balanced accuracy ± 0.008 SE | base · 4206 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_binary / LFM2-2.6B (base) |
NHANES Mortality / Multiclass
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3.5-9B | 0.391 balanced accuracy ± 0.013 SE | L-LLM · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / L-Qwen3.5-9B |
| Gemini-3.1-Pro | 0.380 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Gemini-3.1-Pro |
| GPT-5 | 0.363 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / GPT-5 |
| L-LFM2-2.6B | 0.361 balanced accuracy ± 0.013 SE | L-LLM · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / L-LFM2-2.6B |
| L-LFM2-1.2B | 0.353 balanced accuracy ± 0.013 SE | L-LLM · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / L-LFM2-1.2B |
| GPT-5-mini | 0.352 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / GPT-5-mini |
| Claude-Opus-4.6 | 0.346 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Claude-Opus-4.6 |
| Kimi-K2.5 | 0.335 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Kimi-K2.5 |
| L-Qwen3-0.6B | 0.332 balanced accuracy ± 0.013 SE | L-LLM · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / L-Qwen3-0.6B |
| Claude-Sonnet-4.6 | 0.326 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Claude-Sonnet-4.6 |
| Grok-4.3 | 0.325 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Grok-4.3 |
| GPT-4.1 | 0.319 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / GPT-4.1 |
| L-Qwen3-1.7B | 0.310 balanced accuracy ± 0.013 SE | L-LLM · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / L-Qwen3-1.7B |
| GPT-5.2 | 0.308 balanced accuracy ± 0.013 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / GPT-5.2 |
| Claude-Opus-4.5 | 0.302 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Claude-Opus-4.5 |
| DeepSeek-R1 | 0.296 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / DeepSeek-R1 |
| GPT-4.1-mini | 0.292 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / GPT-4.1-mini |
| Claude-Sonnet-4.5 | 0.284 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Claude-Sonnet-4.5 |
| Qwen3-0.6B (base) | 0.273 balanced accuracy ± 0.012 SE | base · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Qwen3-0.6B (base) |
| Gemini-2.5-Flash | 0.268 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Gemini-2.5-Flash |
| DeepSeek-V3.2 | 0.262 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / DeepSeek-V3.2 |
| Gemini-2.5-Pro | 0.259 balanced accuracy ± 0.012 SE | frontier · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Gemini-2.5-Pro |
| LFM2-1.2B (base) | 0.254 balanced accuracy ± 0.012 SE | base · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / LFM2-1.2B (base) |
| Qwen3.5-9B (base) | 0.225 balanced accuracy ± 0.011 SE | base · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Qwen3.5-9B (base) |
| LFM2-2.6B (base) | 0.219 balanced accuracy ± 0.011 SE | base · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / LFM2-2.6B (base) |
| Qwen3-1.7B (base) | 0.216 balanced accuracy ± 0.011 SE | base · 1363 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_multiclass / Qwen3-1.7B (base) |
NHANES Mortality / Pairwise
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3.5-9B | 0.685 concordance ± 0.007 SE | L-LLM · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / L-Qwen3.5-9B |
| Gemini-3.1-Pro | 0.684 concordance ± 0.007 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Gemini-3.1-Pro |
| GPT-5-mini | 0.672 concordance ± 0.007 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / GPT-5-mini |
| L-LFM2-1.2B | 0.672 concordance ± 0.007 SE | L-LLM · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / L-LFM2-1.2B |
| GPT-5 | 0.666 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / GPT-5 |
| Claude-Opus-4.6 | 0.665 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Claude-Opus-4.6 |
| L-Qwen3-1.7B | 0.660 concordance ± 0.008 SE | L-LLM · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / L-Qwen3-1.7B |
| Grok-4.3 | 0.659 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Grok-4.3 |
| Claude-Sonnet-4.6 | 0.658 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Claude-Sonnet-4.6 |
| Claude-Opus-4.5 | 0.657 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Claude-Opus-4.5 |
| Claude-Sonnet-4.5 | 0.652 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Claude-Sonnet-4.5 |
| GPT-5.2 | 0.645 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / GPT-5.2 |
| Kimi-K2.5 | 0.643 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Kimi-K2.5 |
| Gemini-2.5-Pro | 0.640 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Gemini-2.5-Pro |
| DeepSeek-R1 | 0.639 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / DeepSeek-R1 |
| L-LFM2-2.6B | 0.639 concordance ± 0.008 SE | L-LLM · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / L-LFM2-2.6B |
| L-Qwen3-0.6B | 0.638 concordance ± 0.008 SE | L-LLM · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / L-Qwen3-0.6B |
| GPT-4.1 | 0.637 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / GPT-4.1 |
| Gemini-2.5-Flash | 0.633 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Gemini-2.5-Flash |
| GPT-4.1-mini | 0.610 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / GPT-4.1-mini |
| Qwen3.5-9B (base) | 0.554 concordance ± 0.008 SE | base · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Qwen3.5-9B (base) |
| LFM2-2.6B (base) | 0.540 concordance ± 0.008 SE | base · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / LFM2-2.6B (base) |
| Qwen3-1.7B (base) | 0.529 concordance ± 0.008 SE | base · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Qwen3-1.7B (base) |
| DeepSeek-V3.2 | 0.518 concordance ± 0.008 SE | frontier · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / DeepSeek-V3.2 |
| Qwen3-0.6B (base) | 0.501 concordance ± 0.008 SE | base · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / Qwen3-0.6B (base) |
| LFM2-1.2B (base) | 0.493 concordance ± 0.008 SE | base · 3889 source prompts | Insilico Medicine · LongevityBench nhanes_mortality_pairwise / LFM2-1.2B (base) |
GEO DNAm Age / Multiclass
| Model | Score | Variant | Source |
|---|---|---|---|
| Claude-Opus-4.5 | 0.257 balanced accuracy ± 0.019 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Claude-Opus-4.5 |
| Claude-Opus-4.6 | 0.231 balanced accuracy ± 0.019 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Claude-Opus-4.6 |
| GPT-5.2 | 0.231 balanced accuracy ± 0.019 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / GPT-5.2 |
| Gemini-3.1-Pro | 0.228 balanced accuracy ± 0.019 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Gemini-3.1-Pro |
| Claude-Sonnet-4.6 | 0.198 balanced accuracy ± 0.018 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Claude-Sonnet-4.6 |
| GPT-5-mini | 0.183 balanced accuracy ± 0.018 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / GPT-5-mini |
| L-Qwen3.5-9B | 0.180 balanced accuracy ± 0.017 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / L-Qwen3.5-9B |
| Grok-4.3 | 0.172 balanced accuracy ± 0.017 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Grok-4.3 |
| Kimi-K2.5 | 0.166 balanced accuracy ± 0.017 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Kimi-K2.5 |
| Claude-Sonnet-4.5 | 0.165 balanced accuracy ± 0.017 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Claude-Sonnet-4.5 |
| GPT-5 | 0.165 balanced accuracy ± 0.017 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / GPT-5 |
| GPT-4.1-mini | 0.158 balanced accuracy ± 0.016 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / GPT-4.1-mini |
| DeepSeek-V3.2 | 0.134 balanced accuracy ± 0.015 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / DeepSeek-V3.2 |
| Gemini-2.5-Flash | 0.119 balanced accuracy ± 0.015 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Gemini-2.5-Flash |
| L-LFM2-1.2B | 0.113 balanced accuracy ± 0.014 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / L-LFM2-1.2B |
| L-Qwen3-0.6B | 0.109 balanced accuracy ± 0.014 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / L-Qwen3-0.6B |
| L-LFM2-2.6B | 0.108 balanced accuracy ± 0.014 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / L-LFM2-2.6B |
| DeepSeek-R1 | 0.102 balanced accuracy ± 0.013 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / DeepSeek-R1 |
| GPT-4.1 | 0.098 balanced accuracy ± 0.014 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / GPT-4.1 |
| L-Qwen3-1.7B | 0.090 balanced accuracy ± 0.013 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / L-Qwen3-1.7B |
| LFM2-1.2B (base) | 0.089 balanced accuracy ± 0.013 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / LFM2-1.2B (base) |
| Qwen3-0.6B (base) | 0.089 balanced accuracy ± 0.013 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Qwen3-0.6B (base) |
| Qwen3.5-9B (base) | 0.081 balanced accuracy ± 0.012 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Qwen3.5-9B (base) |
| LFM2-2.6B (base) | 0.073 balanced accuracy ± 0.012 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / LFM2-2.6B (base) |
| Qwen3-1.7B (base) | 0.062 balanced accuracy ± 0.011 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Qwen3-1.7B (base) |
| Gemini-2.5-Pro | 0.048 balanced accuracy ± 0.010 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_multiclass / Gemini-2.5-Pro |
GEO DNAm Age / Pairwise
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3.5-9B | 0.868 concordance ± 0.015 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / L-Qwen3.5-9B |
| L-LFM2-2.6B | 0.726 concordance ± 0.020 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / L-LFM2-2.6B |
| Kimi-K2.5 | 0.685 concordance ± 0.021 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Kimi-K2.5 |
| GPT-5 | 0.654 concordance ± 0.021 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / GPT-5 |
| L-Qwen3-1.7B | 0.646 concordance ± 0.021 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / L-Qwen3-1.7B |
| GPT-5-mini | 0.642 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / GPT-5-mini |
| Claude-Opus-4.6 | 0.628 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Claude-Opus-4.6 |
| Grok-4.3 | 0.628 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Grok-4.3 |
| Gemini-3.1-Pro | 0.614 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Gemini-3.1-Pro |
| Claude-Sonnet-4.6 | 0.608 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Claude-Sonnet-4.6 |
| Gemini-2.5-Pro | 0.606 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Gemini-2.5-Pro |
| GPT-5.2 | 0.604 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / GPT-5.2 |
| LFM2-1.2B (base) | 0.600 concordance ± 0.063 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / LFM2-1.2B (base) |
| Claude-Opus-4.5 | 0.590 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Claude-Opus-4.5 |
| DeepSeek-V3.2 | 0.588 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / DeepSeek-V3.2 |
| Claude-Sonnet-4.5 | 0.584 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Claude-Sonnet-4.5 |
| GPT-4.1-mini | 0.580 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / GPT-4.1-mini |
| L-LFM2-1.2B | 0.578 concordance ± 0.022 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / L-LFM2-1.2B |
| L-Qwen3-0.6B | 0.574 concordance ± 0.022 SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / L-Qwen3-0.6B |
| Gemini-2.5-Flash | 0.572 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Gemini-2.5-Flash |
| Qwen3.5-9B (base) | 0.560 concordance ± 0.022 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Qwen3.5-9B (base) |
| DeepSeek-R1 | 0.538 concordance ± 0.022 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / DeepSeek-R1 |
| GPT-4.1 | 0.518 concordance ± 0.023 SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / GPT-4.1 |
| LFM2-2.6B (base) | 0.512 concordance ± 0.024 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / LFM2-2.6B (base) |
| Qwen3-1.7B (base) | 0.508 concordance ± 0.022 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Qwen3-1.7B (base) |
| Qwen3-0.6B (base) | 0.502 concordance ± 0.022 SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_pairwise / Qwen3-0.6B (base) |
GEO DNAm Age / Regression · lower is better
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3.5-9B | 12.9 yr mean absolute error ± 0.4 yr SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / L-Qwen3.5-9B |
| GPT-4.1-mini | 14.5 yr mean absolute error ± 0.4 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / GPT-4.1-mini |
| Claude-Opus-4.5 | 14.5 yr mean absolute error ± 0.4 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Claude-Opus-4.5 |
| DeepSeek-V3.2 | 15.0 yr mean absolute error ± 0.4 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / DeepSeek-V3.2 |
| Claude-Opus-4.6 | 15.4 yr mean absolute error ± 0.7 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Claude-Opus-4.6 |
| GPT-5.2 | 15.4 yr mean absolute error ± 0.5 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / GPT-5.2 |
| Claude-Sonnet-4.6 | 15.8 yr mean absolute error ± 0.5 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Claude-Sonnet-4.6 |
| Grok-4.3 | 15.9 yr mean absolute error ± 0.5 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Grok-4.3 |
| Gemini-3.1-Pro | 16.2 yr mean absolute error ± 0.6 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Gemini-3.1-Pro |
| GPT-5 | 16.5 yr mean absolute error ± 0.6 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / GPT-5 |
| GPT-5-mini | 17.6 yr mean absolute error ± 0.6 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / GPT-5-mini |
| Claude-Sonnet-4.5 | 18.1 yr mean absolute error ± 0.6 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Claude-Sonnet-4.5 |
| L-Qwen3-0.6B | 18.8 yr mean absolute error ± 0.5 yr SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / L-Qwen3-0.6B |
| L-Qwen3-1.7B | 19.3 yr mean absolute error ± 0.7 yr SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / L-Qwen3-1.7B |
| L-LFM2-2.6B | 20.2 yr mean absolute error ± 0.6 yr SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / L-LFM2-2.6B |
| Kimi-K2.5 | 20.3 yr mean absolute error ± 0.8 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Kimi-K2.5 |
| Gemini-2.5-Flash | 20.5 yr mean absolute error ± 0.8 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Gemini-2.5-Flash |
| DeepSeek-R1 | 21.3 yr mean absolute error ± 0.8 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / DeepSeek-R1 |
| GPT-4.1 | 24.3 yr mean absolute error ± 0.9 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / GPT-4.1 |
| Gemini-2.5-Pro | 24.7 yr mean absolute error ± 0.9 yr SE | frontier · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Gemini-2.5-Pro |
| L-LFM2-1.2B | 27.4 yr mean absolute error ± 0.8 yr SE | L-LLM · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / L-LFM2-1.2B |
| Qwen3.5-9B (base) | 29.4 yr mean absolute error ± 0.8 yr SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Qwen3.5-9B (base) |
| Qwen3-0.6B (base) | 44.7 yr mean absolute error ± 0.9 yr SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Qwen3-0.6B (base) |
| Qwen3-1.7B (base) | 51.4 yr mean absolute error ± 1.1 yr SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / Qwen3-1.7B (base) |
| LFM2-1.2B (base) | 51.6 yr mean absolute error ± 1.1 yr SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / LFM2-1.2B (base) |
| LFM2-2.6B (base) | 65.2 yr mean absolute error ± 5.7 yr SE | base · 500 source prompts | Insilico Medicine · LongevityBench geodnam_age_regression / LFM2-2.6B (base) |
GTEx Transcriptomic Age / Multiclass
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3.5-9B | 0.430 balanced accuracy ± 0.017 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / L-Qwen3.5-9B |
| L-LFM2-2.6B | 0.410 balanced accuracy ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / L-LFM2-2.6B |
| L-Qwen3-1.7B | 0.388 balanced accuracy ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / L-Qwen3-1.7B |
| L-LFM2-1.2B | 0.374 balanced accuracy ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / L-LFM2-1.2B |
| L-Qwen3-0.6B | 0.362 balanced accuracy ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / L-Qwen3-0.6B |
| Gemini-3.1-Pro | 0.360 balanced accuracy ± 0.016 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Gemini-3.1-Pro |
| Claude-Sonnet-4.6 | 0.337 balanced accuracy ± 0.016 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Claude-Sonnet-4.6 |
| Claude-Opus-4.6 | 0.315 balanced accuracy ± 0.015 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Claude-Opus-4.6 |
| Claude-Sonnet-4.5 | 0.308 balanced accuracy ± 0.015 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Claude-Sonnet-4.5 |
| Claude-Opus-4.5 | 0.286 balanced accuracy ± 0.015 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Claude-Opus-4.5 |
| GPT-5-mini | 0.256 balanced accuracy ± 0.015 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / GPT-5-mini |
| GPT-5.2 | 0.252 balanced accuracy ± 0.015 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / GPT-5.2 |
| Gemini-2.5-Pro | 0.247 balanced accuracy ± 0.014 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Gemini-2.5-Pro |
| Qwen3-0.6B (base) | 0.235 balanced accuracy ± 0.014 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Qwen3-0.6B (base) |
| Kimi-K2.5 | 0.218 balanced accuracy ± 0.014 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Kimi-K2.5 |
| DeepSeek-V3.2 | 0.216 balanced accuracy ± 0.014 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / DeepSeek-V3.2 |
| Grok-4.3 | 0.194 balanced accuracy ± 0.013 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Grok-4.3 |
| Qwen3-1.7B (base) | 0.193 balanced accuracy ± 0.013 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Qwen3-1.7B (base) |
| GPT-4.1 | 0.187 balanced accuracy ± 0.013 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / GPT-4.1 |
| DeepSeek-R1 | 0.183 balanced accuracy ± 0.013 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / DeepSeek-R1 |
| LFM2-1.2B (base) | 0.183 balanced accuracy ± 0.013 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / LFM2-1.2B (base) |
| GPT-5 | 0.157 balanced accuracy ± 0.012 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / GPT-5 |
| GPT-4.1-mini | 0.151 balanced accuracy ± 0.012 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / GPT-4.1-mini |
| Gemini-2.5-Flash | 0.144 balanced accuracy ± 0.012 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Gemini-2.5-Flash |
| LFM2-2.6B (base) | 0.115 balanced accuracy ± 0.011 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / LFM2-2.6B (base) |
| Qwen3.5-9B (base) | 0.105 balanced accuracy ± 0.010 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_multiclass / Qwen3.5-9B (base) |
GTEx Transcriptomic Age / Pairwise
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3.5-9B | 0.671 concordance ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / L-Qwen3.5-9B |
| L-Qwen3-1.7B | 0.613 concordance ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / L-Qwen3-1.7B |
| L-Qwen3-0.6B | 0.598 concordance ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / L-Qwen3-0.6B |
| L-LFM2-2.6B | 0.590 concordance ± 0.016 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / L-LFM2-2.6B |
| L-LFM2-1.2B | 0.572 concordance ± 0.017 SE | L-LLM · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / L-LFM2-1.2B |
| Kimi-K2.5 | 0.557 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Kimi-K2.5 |
| GPT-5-mini | 0.544 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / GPT-5-mini |
| Gemini-2.5-Flash | 0.539 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Gemini-2.5-Flash |
| Claude-Sonnet-4.6 | 0.536 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Claude-Sonnet-4.6 |
| Grok-4.3 | 0.534 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Grok-4.3 |
| GPT-5 | 0.532 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / GPT-5 |
| Claude-Opus-4.5 | 0.531 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Claude-Opus-4.5 |
| Claude-Opus-4.6 | 0.531 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Claude-Opus-4.6 |
| Gemini-3.1-Pro | 0.528 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Gemini-3.1-Pro |
| Claude-Sonnet-4.5 | 0.526 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Claude-Sonnet-4.5 |
| Gemini-2.5-Pro | 0.526 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Gemini-2.5-Pro |
| GPT-5.2 | 0.522 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / GPT-5.2 |
| DeepSeek-R1 | 0.519 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / DeepSeek-R1 |
| LFM2-1.2B (base) | 0.514 concordance ± 0.017 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / LFM2-1.2B (base) |
| Qwen3-0.6B (base) | 0.512 concordance ± 0.017 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Qwen3-0.6B (base) |
| DeepSeek-V3.2 | 0.494 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / DeepSeek-V3.2 |
| LFM2-2.6B (base) | 0.494 concordance ± 0.017 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / LFM2-2.6B (base) |
| GPT-4.1 | 0.493 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / GPT-4.1 |
| GPT-4.1-mini | 0.487 concordance ± 0.017 SE | frontier · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / GPT-4.1-mini |
| Qwen3-1.7B (base) | 0.483 concordance ± 0.017 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Qwen3-1.7B (base) |
| Qwen3.5-9B (base) | 0.479 concordance ± 0.017 SE | base · 900 source prompts | Insilico Medicine · LongevityBench gtex_age_pairwise / Qwen3.5-9B (base) |
Olink Proteomic Age / Multiclass
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3-0.6B | 0.454 balanced accuracy ± 0.052 SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / L-Qwen3-0.6B |
| L-Qwen3.5-9B | 0.438 balanced accuracy ± 0.051 SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / L-Qwen3.5-9B |
| L-Qwen3-1.7B | 0.412 balanced accuracy ± 0.051 SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / L-Qwen3-1.7B |
| L-LFM2-2.6B | 0.399 balanced accuracy ± 0.051 SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / L-LFM2-2.6B |
| Claude-Opus-4.5 | 0.359 balanced accuracy ± 0.050 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Claude-Opus-4.5 |
| Claude-Opus-4.6 | 0.345 balanced accuracy ± 0.049 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Claude-Opus-4.6 |
| Claude-Sonnet-4.6 | 0.328 balanced accuracy ± 0.049 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Claude-Sonnet-4.6 |
| GPT-4.1 | 0.294 balanced accuracy ± 0.047 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / GPT-4.1 |
| Qwen3.5-9B (base) | 0.294 balanced accuracy ± 0.047 SE | base | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Qwen3.5-9B (base) |
| Grok-4.3 | 0.289 balanced accuracy ± 0.047 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Grok-4.3 |
| GPT-5.2 | 0.283 balanced accuracy ± 0.047 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / GPT-5.2 |
| L-LFM2-1.2B | 0.277 balanced accuracy ± 0.046 SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / L-LFM2-1.2B |
| DeepSeek-R1 | 0.271 balanced accuracy ± 0.046 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / DeepSeek-R1 |
| LFM2-2.6B (base) | 0.264 balanced accuracy ± 0.046 SE | base | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / LFM2-2.6B (base) |
| LFM2-1.2B (base) | 0.264 balanced accuracy ± 0.046 SE | base | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / LFM2-1.2B (base) |
| GPT-4.1-mini | 0.248 balanced accuracy ± 0.045 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / GPT-4.1-mini |
| Gemini-3.1-Pro | 0.236 balanced accuracy ± 0.044 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Gemini-3.1-Pro |
| Gemini-2.5-Flash | 0.230 balanced accuracy ± 0.044 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Gemini-2.5-Flash |
| Kimi-K2.5 | 0.230 balanced accuracy ± 0.044 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Kimi-K2.5 |
| Qwen3-1.7B (base) | 0.224 balanced accuracy ± 0.052 SE | base | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Qwen3-1.7B (base) |
| Claude-Sonnet-4.5 | 0.219 balanced accuracy ± 0.043 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Claude-Sonnet-4.5 |
| GPT-5-mini | 0.217 balanced accuracy ± 0.043 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / GPT-5-mini |
| Qwen3-0.6B (base) | 0.213 balanced accuracy ± 0.043 SE | base | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Qwen3-0.6B (base) |
| GPT-5 | 0.194 balanced accuracy ± 0.041 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / GPT-5 |
| DeepSeek-V3.2 | 0.188 balanced accuracy ± 0.041 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / DeepSeek-V3.2 |
| Gemini-2.5-Pro | 0.187 balanced accuracy ± 0.040 SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_multiclass / Gemini-2.5-Pro |
Olink Proteomic Age / Pairwise (multi-cohort)
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3-0.6B | 0.824 concordance ± 0.009 SE | L-LLM · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / L-Qwen3-0.6B |
| L-Qwen3-1.7B | 0.824 concordance ± 0.009 SE | L-LLM · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / L-Qwen3-1.7B |
| L-Qwen3.5-9B | 0.810 concordance ± 0.010 SE | L-LLM · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / L-Qwen3.5-9B |
| L-LFM2-2.6B | 0.800 concordance ± 0.010 SE | L-LLM · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / L-LFM2-2.6B |
| L-LFM2-1.2B | 0.771 concordance ± 0.011 SE | L-LLM · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / L-LFM2-1.2B |
| Gemini-3.1-Pro | 0.703 concordance ± 0.011 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Gemini-3.1-Pro |
| GPT-5 | 0.649 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / GPT-5 |
| Claude-Opus-4.6 | 0.645 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Claude-Opus-4.6 |
| Grok-4.3 | 0.645 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Grok-4.3 |
| Kimi-K2.5 | 0.641 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Kimi-K2.5 |
| DeepSeek-R1 | 0.640 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / DeepSeek-R1 |
| Claude-Sonnet-4.5 | 0.639 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Claude-Sonnet-4.5 |
| GPT-5-mini | 0.636 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / GPT-5-mini |
| Claude-Sonnet-4.6 | 0.620 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Claude-Sonnet-4.6 |
| Claude-Opus-4.5 | 0.618 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Claude-Opus-4.5 |
| GPT-5.2 | 0.616 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / GPT-5.2 |
| Gemini-2.5-Pro | 0.612 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Gemini-2.5-Pro |
| DeepSeek-V3.2 | 0.581 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / DeepSeek-V3.2 |
| Gemini-2.5-Flash | 0.581 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Gemini-2.5-Flash |
| GPT-4.1-mini | 0.536 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / GPT-4.1-mini |
| LFM2-2.6B (base) | 0.521 concordance ± 0.012 SE | base · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / LFM2-2.6B (base) |
| Qwen3-1.7B (base) | 0.518 concordance ± 0.012 SE | base · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Qwen3-1.7B (base) |
| LFM2-1.2B (base) | 0.506 concordance ± 0.013 SE | base · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / LFM2-1.2B (base) |
| GPT-4.1 | 0.500 concordance ± 0.012 SE | frontier · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / GPT-4.1 |
| Qwen3.5-9B (base) | 0.496 concordance ± 0.012 SE | base · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Qwen3.5-9B (base) |
| Qwen3-0.6B (base) | 0.478 concordance ± 0.012 SE | base · 1615 source prompts | Insilico Medicine · LongevityBench olink_age_pairwise / Qwen3-0.6B (base) |
Olink Proteomic Age / Regression · lower is better
| Model | Score | Variant | Source |
|---|---|---|---|
| L-Qwen3-0.6B | 5.7 yr mean absolute error ± 0.5 yr SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_regression / L-Qwen3-0.6B |
| L-Qwen3.5-9B | 5.8 yr mean absolute error ± 0.5 yr SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_regression / L-Qwen3.5-9B |
| L-Qwen3-1.7B | 6.2 yr mean absolute error ± 0.5 yr SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_regression / L-Qwen3-1.7B |
| L-LFM2-2.6B | 7.2 yr mean absolute error ± 0.6 yr SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_regression / L-LFM2-2.6B |
| L-LFM2-1.2B | 8.5 yr mean absolute error ± 0.7 yr SE | L-LLM | Insilico Medicine · LongevityBench olink_pooled_age_regression / L-LFM2-1.2B |
| Kimi-K2.5 | 10.1 yr mean absolute error ± 0.8 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Kimi-K2.5 |
| GPT-4.1-mini | 10.4 yr mean absolute error ± 0.8 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / GPT-4.1-mini |
| DeepSeek-R1 | 10.7 yr mean absolute error ± 1.0 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / DeepSeek-R1 |
| GPT-4.1 | 10.8 yr mean absolute error ± 0.9 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / GPT-4.1 |
| Claude-Sonnet-4.6 | 10.8 yr mean absolute error ± 0.9 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Claude-Sonnet-4.6 |
| Qwen3.5-9B (base) | 10.9 yr mean absolute error ± 0.7 yr SE | base | Insilico Medicine · LongevityBench olink_pooled_age_regression / Qwen3.5-9B (base) |
| Grok-4.3 | 11.1 yr mean absolute error ± 0.9 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Grok-4.3 |
| Gemini-2.5-Flash | 11.5 yr mean absolute error ± 1.0 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Gemini-2.5-Flash |
| LFM2-2.6B (base) | 11.7 yr mean absolute error ± 0.7 yr SE | base | Insilico Medicine · LongevityBench olink_pooled_age_regression / LFM2-2.6B (base) |
| GPT-5.2 | 13.5 yr mean absolute error ± 1.0 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / GPT-5.2 |
| Gemini-2.5-Pro | 13.8 yr mean absolute error ± 0.9 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Gemini-2.5-Pro |
| GPT-5-mini | 13.9 yr mean absolute error ± 1.0 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / GPT-5-mini |
| DeepSeek-V3.2 | 16.5 yr mean absolute error ± 1.3 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / DeepSeek-V3.2 |
| GPT-5 | 17.2 yr mean absolute error ± 1.2 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / GPT-5 |
| LFM2-1.2B (base) | 17.2 yr mean absolute error ± 3.6 yr SE | base | Insilico Medicine · LongevityBench olink_pooled_age_regression / LFM2-1.2B (base) |
| Gemini-3.1-Pro | 17.6 yr mean absolute error ± 1.2 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Gemini-3.1-Pro |
| Claude-Opus-4.6 | 17.9 yr mean absolute error ± 0.9 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Claude-Opus-4.6 |
| Claude-Opus-4.5 | 22.7 yr mean absolute error ± 3.9 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Claude-Opus-4.5 |
| Qwen3-1.7B (base) | 29.0 yr mean absolute error ± 1.5 yr SE | base | Insilico Medicine · LongevityBench olink_pooled_age_regression / Qwen3-1.7B (base) |
| Qwen3-0.6B (base) | 41.8 yr mean absolute error ± 1.5 yr SE | base | Insilico Medicine · LongevityBench olink_pooled_age_regression / Qwen3-0.6B (base) |
| Claude-Sonnet-4.5 | 52.6 yr mean absolute error ± 11.3 yr SE | frontier | Insilico Medicine · LongevityBench olink_pooled_age_regression / Claude-Sonnet-4.5 |
OpenGenes Expression / Binary (masked)
| Model | Score | Variant | Source |
|---|---|---|---|
| L-LFM2-1.2B | 0.780 balanced accuracy ± 0.023 SE | L-LLM | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / L-LFM2-1.2B |
| L-LFM2-2.6B | 0.774 balanced accuracy ± 0.023 SE | L-LLM | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / L-LFM2-2.6B |
| L-Qwen3-0.6B | 0.755 balanced accuracy ± 0.024 SE | L-LLM | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / L-Qwen3-0.6B |
| L-Qwen3-1.7B | 0.744 balanced accuracy ± 0.024 SE | L-LLM | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / L-Qwen3-1.7B |
| L-Qwen3.5-9B | 0.710 balanced accuracy ± 0.025 SE | L-LLM | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / L-Qwen3.5-9B |
| Gemini-3.1-Pro | 0.677 balanced accuracy ± 0.026 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Gemini-3.1-Pro |
| GPT-5 | 0.668 balanced accuracy ± 0.026 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / GPT-5 |
| Claude-Opus-4.6 | 0.662 balanced accuracy ± 0.026 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Claude-Opus-4.6 |
| Claude-Sonnet-4.6 | 0.652 balanced accuracy ± 0.026 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Claude-Sonnet-4.6 |
| Grok-4.3 | 0.643 balanced accuracy ± 0.026 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Grok-4.3 |
| GPT-5-mini | 0.625 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / GPT-5-mini |
| DeepSeek-R1 | 0.622 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / DeepSeek-R1 |
| Kimi-K2.5 | 0.616 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Kimi-K2.5 |
| Gemini-2.5-Pro | 0.598 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Gemini-2.5-Pro |
| Claude-Opus-4.5 | 0.598 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Claude-Opus-4.5 |
| Claude-Sonnet-4.5 | 0.593 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Claude-Sonnet-4.5 |
| LFM2-1.2B (base) | 0.585 balanced accuracy ± 0.029 SE | base | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / LFM2-1.2B (base) |
| Gemini-2.5-Flash | 0.577 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Gemini-2.5-Flash |
| Qwen3-1.7B (base) | 0.552 balanced accuracy ± 0.027 SE | base | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Qwen3-1.7B (base) |
| GPT-5.2 | 0.503 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / GPT-5.2 |
| Qwen3-0.6B (base) | 0.501 balanced accuracy ± 0.027 SE | base | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Qwen3-0.6B (base) |
| LFM2-2.6B (base) | 0.482 balanced accuracy ± 0.027 SE | base | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / LFM2-2.6B (base) |
| Qwen3.5-9B (base) | 0.478 balanced accuracy ± 0.027 SE | base | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / Qwen3.5-9B (base) |
| GPT-4.1 | 0.477 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / GPT-4.1 |
| DeepSeek-V3.2 | 0.467 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / DeepSeek-V3.2 |
| GPT-4.1-mini | 0.453 balanced accuracy ± 0.027 SE | frontier | Insilico Medicine · LongevityBench opengenes_expression_binary_masked / GPT-4.1-mini |
SynergyAge Lifespan / Pairwise
| Model | Score | Variant | Source |
|---|---|---|---|
| Claude-Opus-4.5 | 0.846 concordance ± 0.071 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Claude-Opus-4.5 |
| GPT-4.1 | 0.846 concordance ± 0.071 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / GPT-4.1 |
| GPT-5-mini | 0.846 concordance ± 0.071 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / GPT-5-mini |
| Gemini-3.1-Pro | 0.846 concordance ± 0.071 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Gemini-3.1-Pro |
| Kimi-K2.5 | 0.833 concordance ± 0.076 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Kimi-K2.5 |
| Claude-Sonnet-4.5 | 0.808 concordance ± 0.077 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Claude-Sonnet-4.5 |
| Claude-Sonnet-4.6 | 0.808 concordance ± 0.077 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Claude-Sonnet-4.6 |
| DeepSeek-R1 | 0.808 concordance ± 0.077 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / DeepSeek-R1 |
| GPT-5 | 0.808 concordance ± 0.077 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / GPT-5 |
| GPT-5.2 | 0.808 concordance ± 0.077 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / GPT-5.2 |
| Grok-4.3 | 0.769 concordance ± 0.083 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Grok-4.3 |
| L-Qwen3-1.7B | 0.769 concordance ± 0.083 SE | L-LLM | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / L-Qwen3-1.7B |
| DeepSeek-V3.2 | 0.731 concordance ± 0.087 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / DeepSeek-V3.2 |
| GPT-4.1-mini | 0.731 concordance ± 0.087 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / GPT-4.1-mini |
| L-Qwen3.5-9B | 0.731 concordance ± 0.087 SE | L-LLM | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / L-Qwen3.5-9B |
| Claude-Opus-4.6 | 0.654 concordance ± 0.093 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Claude-Opus-4.6 |
| Gemini-2.5-Flash | 0.654 concordance ± 0.093 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Gemini-2.5-Flash |
| Gemini-2.5-Pro | 0.654 concordance ± 0.093 SE | frontier | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Gemini-2.5-Pro |
| L-Qwen3-0.6B | 0.654 concordance ± 0.093 SE | L-LLM | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / L-Qwen3-0.6B |
| LFM2-1.2B (base) | 0.615 concordance ± 0.095 SE | base | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / LFM2-1.2B (base) |
| LFM2-2.6B (base) | 0.615 concordance ± 0.095 SE | base | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / LFM2-2.6B (base) |
| L-LFM2-2.6B | 0.577 concordance ± 0.097 SE | L-LLM | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / L-LFM2-2.6B |
| Qwen3.5-9B (base) | 0.577 concordance ± 0.097 SE | base | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Qwen3.5-9B (base) |
| L-LFM2-1.2B | 0.538 concordance ± 0.098 SE | L-LLM | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / L-LFM2-1.2B |
| Qwen3-1.7B (base) | 0.462 concordance ± 0.098 SE | base | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Qwen3-1.7B (base) |
| Qwen3-0.6B (base) | 0.346 concordance ± 0.093 SE | base | Insilico Medicine · LongevityBench synergyage_lifespan_pairwise_minimal / Qwen3-0.6B (base) |
SynergyAge Lifespan / Regression · lower is better
| Model | Score | Variant | Source |
|---|---|---|---|
| Claude-Opus-4.5 | 12.0 yr mean absolute error ± 2.4 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Claude-Opus-4.5 |
| L-Qwen3.5-9B | 12.4 yr mean absolute error ± 2.3 yr SE | L-LLM · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / L-Qwen3.5-9B |
| L-LFM2-2.6B | 13.7 yr mean absolute error ± 2.5 yr SE | L-LLM · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / L-LFM2-2.6B |
| Claude-Opus-4.6 | 13.8 yr mean absolute error ± 2.4 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Claude-Opus-4.6 |
| Gemini-3.1-Pro | 15.1 yr mean absolute error ± 2.6 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Gemini-3.1-Pro |
| GPT-5.2 | 15.6 yr mean absolute error ± 1.7 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / GPT-5.2 |
| Kimi-K2.5 | 15.8 yr mean absolute error ± 3.6 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Kimi-K2.5 |
| GPT-5-mini | 17.2 yr mean absolute error ± 3.6 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / GPT-5-mini |
| L-LFM2-1.2B | 17.7 yr mean absolute error ± 3.0 yr SE | L-LLM · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / L-LFM2-1.2B |
| DeepSeek-R1 | 18.0 yr mean absolute error ± 3.6 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / DeepSeek-R1 |
| L-Qwen3-1.7B | 18.0 yr mean absolute error ± 3.4 yr SE | L-LLM · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / L-Qwen3-1.7B |
| GPT-4.1-mini | 18.1 yr mean absolute error ± 4.4 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / GPT-4.1-mini |
| Claude-Sonnet-4.5 | 18.2 yr mean absolute error ± 3.1 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Claude-Sonnet-4.5 |
| GPT-5 | 19.2 yr mean absolute error ± 3.5 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / GPT-5 |
| GPT-4.1 | 19.2 yr mean absolute error ± 5.1 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / GPT-4.1 |
| Claude-Sonnet-4.6 | 19.5 yr mean absolute error ± 3.4 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Claude-Sonnet-4.6 |
| Grok-4.3 | 20.1 yr mean absolute error ± 3.4 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Grok-4.3 |
| Gemini-2.5-Flash | 21.1 yr mean absolute error ± 3.7 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Gemini-2.5-Flash |
| DeepSeek-V3.2 | 21.3 yr mean absolute error ± 5.1 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / DeepSeek-V3.2 |
| Gemini-2.5-Pro | 22.1 yr mean absolute error ± 4.1 yr SE | frontier · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Gemini-2.5-Pro |
| Qwen3.5-9B (base) | 23.0 yr mean absolute error ± 4.1 yr SE | base · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Qwen3.5-9B (base) |
| LFM2-2.6B (base) | 24.7 yr mean absolute error ± 3.4 yr SE | base · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / LFM2-2.6B (base) |
| Qwen3-1.7B (base) | 24.8 yr mean absolute error ± 4.3 yr SE | base · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Qwen3-1.7B (base) |
| Qwen3-0.6B (base) | 25.2 yr mean absolute error ± 4.9 yr SE | base · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / Qwen3-0.6B (base) |
| L-Qwen3-0.6B | 29.6 yr mean absolute error ± 7.9 yr SE | L-LLM · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / L-Qwen3-0.6B |
| LFM2-1.2B (base) | 44.8 yr mean absolute error ± 8.8 yr SE | base · 26 source prompts | Insilico Medicine · LongevityBench synergyage_lifespan_regression / LFM2-1.2B (base) |
Sources
- external Insilico Medicine · LongevityBench