Benchmark Almanac

TxBench-Preclinical-Pharmacology

Gemini 4 Carbon 66% leads since October 2026
Not in model cards

Agents analyze preclinical pharmacology data and make development decisions. A grader checks each decision.

Part of LatchBio's BioBench suite on benchmarks.bio. The V1 set has 100 evaluations. Release month is the GitHub repo creation month.

Made by
LatchBio
Released
June 2026
Kind
Genomics and life sciences
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Strongest model tested
Gemini 4 Carbon #2 on the capability index
Last model added
November 2026
Scores refreshed
November 13th, 2026
30
Models scored on its own leaderboard

Gemini 4 Carbon leads with 66%

Best score per model from the benchmark's own leaderboard, read on November 13th, 2026. Where a model ran with several settings or harnesses, we keep its best.

30 models. Scroll sideways for the rest.

  1. 66.0 Gemini 4 Carbon
  2. 65.3 Claude Opus 5.5
  3. 64.7 Gemini 4 Argon
  4. 64.3 Claude Sonnet 5.5
  5. 63.0 Claude Opus 5
  6. 63.0 GPT-6 Sol
  7. 61.7 Grok 4.6
  8. 61.3 GPT-5.6 Sol
  9. 61.3 Gemini 4 Hydrogen
  10. 59.7 Claude Fable 5.1
  11. 59.7 GPT-6.1 Sol
  12. 59.3 GPT-6 Astra
  13. 57.7 Claude Opus 4.8
  14. 55.7 GPT-6.1 Terra
  15. 55.3 GPT-5.5
  16. 55.3 Gemini 3.7 Flash
  17. 54.7 Gemini 3.8 Flash
  18. 54.7 Grok 4.7
  19. 49.3 Claude Sonnet 5
  20. 46.7 DeepSeek V4.1 Flash
  21. 45.3 Claude Opus 4.7
  22. 44.3 GPT-6 Luna
  23. 43.7 GPT-5.6 Luna
  24. 42.3 Mistral Large 4
  25. 39.7 Kimi K3
  26. 39.7 Beam 1.0
  27. 35.0 GLM-5.3
  28. 34.3 Claude Sonnet 4.6
  29. 22.7 Nemotron 3 Ultra
  30. 10.3 Nemotron 3.5 Lightning 30B-A3B
  1. 1 Gemini 4 Carbon 66%
  2. 2 Claude Opus 5.5 LSVP 65.3%
  3. 3 Gemini 4 Argon 64.7%
  4. 4 Claude Sonnet 5.5 64.3%
  5. 5 Claude Opus 5 63%
  6. 6 GPT-6 Sol 63%
  7. 7 Grok 4.6 61.7%
  8. 8 GPT-5.6 Sol 61.3%
  9. 9 Gemini 4 Hydrogen 61.3%
  10. 10 Claude Fable 5.1 LSVP 59.7%
  11. 11 GPT-6.1 Sol 59.7%
  12. 12 GPT-6 Astra 59.3%
  13. 13 Claude Opus 4.8 57.7%
  14. 14 GPT-6.1 Terra 55.7%
  15. 15 GPT-5.5 55.3%
  16. 16 Gemini 3.7 Flash 55.3%
  17. 17 Gemini 3.8 Flash 54.7%
  18. 18 Grok 4.7 54.7%
  19. 19 Claude Sonnet 5 49.3%
  20. 20 DeepSeek V4.1 Flash 46.7%
  21. 21 Claude Opus 4.7 45.3%
  22. 22 GPT-6 Luna 44.3%
  23. 23 GPT-5.6 Luna 43.7%
  24. 24 Mistral Large 4 42.3%
  25. 25 Kimi K3 39.7%
  26. 26 Beam 1.0 39.7%
  27. 27 GLM-5.3 35%
  28. 28 Claude Sonnet 4.6 34.3%
  29. 29 Nemotron 3 Ultra 22.7%
  30. 30 Nemotron 3.5 Lightning 30B-A3B 10.3%

Source: Official TxBench-Preclinical-Pharmacology leaderboard

#ModelScoreSettingDateSource
1 Gemini 4 Carbon 66% Pi Oct 2026 Official leaderboard
2 Claude Opus 5.5 65.3% LSVP, Pi Sep 2026 Official leaderboard
3 Gemini 4 Argon 64.7% Pi Sep 2026 Official leaderboard
4 Claude Sonnet 5.5 64.3% Pi Sep 2026 Official leaderboard
5 Claude Opus 5 63% Pi Jul 2026 Official leaderboard
6 GPT-6 Sol 63% Pi Sep 2026 Official leaderboard
7 Grok 4.6 61.7% Pi Aug 2026 Official leaderboard
8 GPT-5.6 Sol 61.3% Pi Jul 2026 Official leaderboard
9 Gemini 4 Hydrogen 61.3% Pi Nov 2026 Official leaderboard
10 Claude Fable 5.1 59.7% LSVP, Pi Sep 2026 Official leaderboard
11 GPT-6.1 Sol 59.7% Pi Sep 2026 Official leaderboard
12 GPT-6 Astra 59.3% Pi Sep 2026 Official leaderboard
13 Claude Opus 4.8 57.7% Pi May 2026 Official leaderboard
14 GPT-6.1 Terra 55.7% Pi Oct 2026 Official leaderboard
15 GPT-5.5 55.3% Pi Apr 2026 Official leaderboard
16 Gemini 3.7 Flash 55.3% Pi Aug 2026 Official leaderboard
17 Gemini 3.8 Flash 54.7% Pi Sep 2026 Official leaderboard
18 Grok 4.7 54.7% Pi Sep 2026 Official leaderboard
19 Claude Sonnet 5 49.3% Pi Jun 2026 Official leaderboard
20 DeepSeek V4.1 Flash 46.7% Pi Sep 2026 Official leaderboard
21 Claude Opus 4.7 45.3% Pi Apr 2026 Official leaderboard
22 GPT-6 Luna 44.3% Pi Sep 2026 Official leaderboard
23 GPT-5.6 Luna 43.7% Codex Jul 2026 Official leaderboard
24 Mistral Large 4 42.3% Pi Oct 2026 Official leaderboard
25 Kimi K3 39.7% Pi Jul 2026 Official leaderboard
26 Beam 1.0 39.7% Pi Oct 2026 Official leaderboard
27 GLM-5.3 35% Pi Aug 2026 Official leaderboard
28 Claude Sonnet 4.6 34.3% Pi Feb 2026 Official leaderboard
29 Nemotron 3 Ultra 22.7% Pi Jun 2026 Official leaderboard
30 Nemotron 3.5 Lightning 30B-A3B 10.3% Pi Aug 2026 Official leaderboard

5 models have held number one on the official leaderboard since its release in June 2026

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

Anthropic OpenAI Google DeepMind
Own leaderboard 5 leaders
Gemini 4 Carbon 66%

Source: Official TxBench-Preclinical-Pharmacology 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 Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Carbon Claude Opus 5 Claude Opus 4.8 Claude Opus 4.7 Claude Sonnet 4.6 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Carbon Claude Opus 4.8

Source: Official TxBench-Preclinical-Pharmacology leaderboard

No lab model card reports it yet

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

More in Biology and chemistry

All 109 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%