Benchmark Almanac

BioSecBench-Function

Claude Opus 5.5 73.6% leads since September 2026
Only in system cards

Agents recover biological function from experimental data. Scored by verifiable checks. One of three LatchBio BioSecBench evaluations.

Repo is Apache 2.0 with a public subset. Scored on benchmarks.bio V1 as a capability benchmark. Paper link pending.

Made by
LatchBio
Released
September 2026
Kind
Genomics and life sciences
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Strongest model tested
GPT-6 Astra #3 on the capability index
Last in a system card
September 2026
Last model added
October 2026
Scores refreshed
November 12th, 2026
27
Models scored on its own leaderboard
1
System card reports it, from 1 lab

Claude Opus 5.5 leads with 73.6%

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

27 models. Scroll sideways for the rest.

  1. 73.6 Claude Opus 5.5
  2. 53.5 Claude Sonnet 5.5
  3. 49.0 Claude Fable 5.1
  4. 48.6 GPT-6 Astra
  5. 46.0 GPT-6.1 Sol
  6. 44.8 Claude Opus 5
  7. 44.4 Gemini 4 Argon
  8. 44.0 Grok 4.6
  9. 43.3 Grok 4.7
  10. 43.1 Claude Opus 4.8
  11. 40.8 DeepSeek V4.1 Flash
  12. 39.3 GPT-6 Sol
  13. 39.0 GPT-6.1 Terra
  14. 36.0 Gemini 3.8 Flash
  15. 35.5 GPT-5.6 Sol
  16. 33.8 Claude Sonnet 5
  17. 33.7 Gemini 3.7 Flash
  18. 32.7 Kimi K3
  19. 31.8 GLM-5.3
  20. 30.8 Claude Opus 4.7
  21. 28.8 Mistral Large 4
  22. 28.3 GPT-5.5
  23. 27.6 GPT-5.6 Luna
  24. 27.3 Claude Sonnet 4.6
  25. 27.3 GPT-6 Luna
  26. 21.6 Nemotron 3 Ultra
  27. 13.8 Nemotron 3.5 Lightning 30B-A3B
  1. 1 Claude Opus 5.5 73.6%
  2. 2 Claude Sonnet 5.5 53.5%
  3. 3 Claude Fable 5.1 LSVP 49%
  4. 4 GPT-6 Astra 48.6%
  5. 5 GPT-6.1 Sol 46%
  6. 6 Claude Opus 5 44.8%
  7. 7 Gemini 4 Argon 44.4%
  8. 8 Grok 4.6 44%
  9. 9 Grok 4.7 43.3%
  10. 10 Claude Opus 4.8 43.1%
  11. 11 DeepSeek V4.1 Flash 40.8%
  12. 12 GPT-6 Sol 39.3%
  13. 13 GPT-6.1 Terra 39%
  14. 14 Gemini 3.8 Flash 36%
  15. 15 GPT-5.6 Sol 35.5%
  16. 16 Claude Sonnet 5 33.8%
  17. 17 Gemini 3.7 Flash 33.7%
  18. 18 Kimi K3 32.7%
  19. 19 GLM-5.3 31.8%
  20. 20 Claude Opus 4.7 30.8%
  21. 21 Mistral Large 4 28.8%
  22. 22 GPT-5.5 28.3%
  23. 23 GPT-5.6 Luna 27.6%
  24. 24 Claude Sonnet 4.6 27.3%
  25. 25 GPT-6 Luna 27.3%
  26. 26 Nemotron 3 Ultra 21.6%
  27. 27 Nemotron 3.5 Lightning 30B-A3B 13.8%

Source: Official BioSecBench-Function leaderboard

#ModelScoreSettingDateSource
1 Claude Opus 5.5 73.6% Pi Sep 2026 Official leaderboard
2 Claude Sonnet 5.5 53.5% Claude Code Sep 2026 Official leaderboard
3 Claude Fable 5.1 49% LSVP, Pi Sep 2026 Official leaderboard
4 GPT-6 Astra 48.6% Pi Sep 2026 Official leaderboard
5 GPT-6.1 Sol 46% Pi Sep 2026 Official leaderboard
6 Claude Opus 5 44.8% Claude Code Jul 2026 Official leaderboard
7 Gemini 4 Argon 44.4% Pi Sep 2026 Official leaderboard
8 Grok 4.6 44% Pi Aug 2026 Official leaderboard
9 Grok 4.7 43.3% Pi Sep 2026 Official leaderboard
10 Claude Opus 4.8 43.1% Pi May 2026 Official leaderboard
11 DeepSeek V4.1 Flash 40.8% Pi Sep 2026 Official leaderboard
12 GPT-6 Sol 39.3% Pi Sep 2026 Official leaderboard
13 GPT-6.1 Terra 39% Pi Oct 2026 Official leaderboard
14 Gemini 3.8 Flash 36% Pi Sep 2026 Official leaderboard
15 GPT-5.6 Sol 35.5% Codex Jul 2026 Official leaderboard
16 Claude Sonnet 5 33.8% Claude Code Jun 2026 Official leaderboard
17 Gemini 3.7 Flash 33.7% Pi Aug 2026 Official leaderboard
18 Kimi K3 32.7% Pi Jul 2026 Official leaderboard
19 GLM-5.3 31.8% Pi Aug 2026 Official leaderboard
20 Claude Opus 4.7 30.8% Pi Apr 2026 Official leaderboard
21 Mistral Large 4 28.8% Pi Oct 2026 Official leaderboard
22 GPT-5.5 28.3% Pi Apr 2026 Official leaderboard
23 GPT-5.6 Luna 27.6% Pi Jul 2026 Official leaderboard
24 Claude Sonnet 4.6 27.3% Claude Code Feb 2026 Official leaderboard
25 GPT-6 Luna 27.3% Pi Sep 2026 Official leaderboard
26 Nemotron 3 Ultra 21.6% Pi Jun 2026 Official leaderboard
27 Nemotron 3.5 Lightning 30B-A3B 13.8% Pi Aug 2026 Official leaderboard

Labs report scores for 1 model in their cards. We plot every source in the charts below.

2 models have held number one on the official leaderboard since its release in September 2026

Lab claims count from the day the card came out, independent runs from the model’s release.

Anthropic
Own leaderboard 2 leaders
Claude Opus 5.5 73.6%

Source: Official BioSecBench-Function leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1 Claude Opus 5 Claude Opus 4.7 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5.1

Sources: Official BioSecBench-Function leaderboard and labs’ model cards

Labs report it only in system cards

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark. Some system cards do, below.

One system card reports it

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Sep 21, 2026 Grok 4.7 model card 43.3 BioSecBench Function, endpoint pass rate, xhigh, Grok Build harness (7.1 table) Bio and chem risk

One lab claim next to the leaderboard

Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. 1 of them match, which usually means one side copied the other. Every benchmark.

Model Lab says Board says Gap
Grok 4.7 43.3% BioSecBench Function, endpoint pass rate, xhigh, Grok Build harness (7.1 table) 43.3% Leaderboard, Pi Same

More in Biology and chemistry

All 108 in Biology and chemistry
  • GeneBench Pro In use Agents analyze messy simulated genomics and biology datasets to estimate a decision-relevant quantity across 129 problems. Scored automatically against known answers. GPT-5.6 Sol 31.5%
  • SUPERChem In use Models solve 500 expert-written chemistry reasoning problems, offered in multimodal and text-only forms. Expert solution paths also allow a reasoning path score. Tencent Hy4 preview 66.4%
  • LifeSciBench In use Models complete 750 expert-written life science research tasks, often with attached figures, tables, or sequence files. Graded against expert rubrics as score and pass rate. GPT-Rosalind-6 61.3%
  • MedChemBench (OpenAI) In use Models answer chemistry questions from an OpenAI internal set, grouped with OpenAI's life science research evaluations. OpenAI has not published the task format. GPT-Rosalind-6 49.6%
  • ProtocolQA Open-Ended In use Models correct errors in published lab protocols using open-ended questions adapted from LAB-Bench ProtocolQA. Scored as pass@1. GPT-6 Cyber 47.2%
  • MolecularIQ In use Models answer questions about molecular graphs, such as counting, indexing and constrained generation, and a symbolic checker verifies each answer. gpt-oss-120b 47.5%