Benchmark Almanac

GeneBench Pro

GPT-5.6 Sol 31.5% leads since July 2026
GPT-Rosalind-6 37.7% leads since October 2026
In use

Agents analyze messy simulated genomics and biology datasets to estimate a decision-relevant quantity across 129 problems. Scored automatically against known answers.

129 problems in 10 domains and 21 sub-domains. Synthetic noisy datasets graded against ground truth. OpenAI calls it a harder expansion of GeneBench. Self-reported scores only so far.

Made by
OpenAI
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
GPT-6 Astra #3 on the capability index
Last in a launch post
September 2026
Last model added
October 2026
Scores refreshed
November 12th, 2026
12
Models scored on its own leaderboard
4
Launch posts list it, from 1 lab
1
System card reports it, from 1 lab

GPT-5.6 Sol leads with 31.5% among models released by July 2026

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. Every run is in the next chart.

GPT-Rosalind-6 leads with 37.7%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

The newest model with a score here came out in July 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 31.5 GPT-5.6 Sol
  2. 28.5 GPT-5.6 Terra
  3. 23.6 GPT-5.6 Luna
  4. 20.5 GPT-5.5
  5. 16.3 GPT-5.4
  6. 16.0 Claude Opus 4.8
  7. 8.53 GPT-5.2
  8. 8.14 Gemini 3.5 Flash
  9. 4.64 GLM-5.2
  10. 3.10 Gemini 3.1 Pro
  11. 2.42 DeepSeek-V4-Pro
  12. 1.47 Grok 4.3
  1. 1 GPT-5.6 Sol pro 31.5%
  2. 2 GPT-5.6 Terra pro 28.5%
  3. 3 GPT-5.6 Luna pro 23.6%
  4. 4 GPT-5.5 pro 20.5%
  5. 5 GPT-5.4 pro 16.3%
  6. 6 Claude Opus 4.8 max available reasoning 16%
  7. 7 GPT-5.2 pro 8.53%
  8. 8 Gemini 3.5 Flash max available reasoning 8.14%
  9. 9 GLM-5.2 max available reasoning 4.64%
  10. 10 Gemini 3.1 Pro max available reasoning 3.10%
  11. 11 DeepSeek-V4-Pro max available reasoning 2.42%
  12. 12 Grok 4.3 max available reasoning 1.47%

Source: Official GeneBench Pro leaderboard

Sources: Labs’ model cards and official GeneBench Pro leaderboard

#ModelScoreSettingDateSource
1 GPT-5.6 Sol 31.5% pro Jul 2026 Official leaderboard
2 GPT-5.6 Terra 28.5% pro Jul 2026 Official leaderboard
3 GPT-5.6 Luna 23.6% pro Jul 2026 Official leaderboard
4 GPT-5.5 20.5% pro Apr 2026 Official leaderboard
5 GPT-5.4 16.3% pro Mar 2026 Official leaderboard
6 Claude Opus 4.8 16% max available reasoning May 2026 Official leaderboard
7 GPT-5.2 8.53% pro Dec 2025 Official leaderboard
8 Gemini 3.5 Flash 8.14% max available reasoning May 2026 Official leaderboard
9 GLM-5.2 4.64% max available reasoning Jun 2026 Official leaderboard
10 Gemini 3.1 Pro 3.10% max available reasoning Feb 2026 Official leaderboard
11 DeepSeek-V4-Pro 2.42% max available reasoning Apr 2026 Official leaderboard
12 Grok 4.3 1.47% max available reasoning Apr 2026 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-Rosalind-6 37.7% Oct 2026 Lab model card
2 GPT-6 Astra 37.1% Sep 2026 Lab model card
3 GPT-5.6 Sol 31.5% pro Jul 2026 Official leaderboard
4 GPT-5.6 Terra 28.5% pro Jul 2026 Official leaderboard
5 GPT-5.6 Luna 23.6% pro Jul 2026 Official leaderboard
6 GPT-5.5 20.5% pro Apr 2026 Official leaderboard
7 GPT-5.4 16.3% pro Mar 2026 Official leaderboard
8 Claude Opus 4.8 16% max available reasoning May 2026 Official leaderboard
9 GPT-5.2 8.53% pro Dec 2025 Official leaderboard
10 Gemini 3.5 Flash 8.14% max available reasoning May 2026 Official leaderboard
11 GLM-5.2 4.64% max available reasoning Jun 2026 Official leaderboard
12 Gemini 3.1 Pro 3.10% max available reasoning Feb 2026 Official leaderboard
13 DeepSeek-V4-Pro 2.42% max available reasoning Apr 2026 Official leaderboard
14 Grok 4.3 1.47% max available reasoning Apr 2026 Official leaderboard

Labs report scores for 5 models in their cards. Switch the scores above to rank them together.

GeneBench-Pro: Test-time compute scaling

The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0% 10% 20% 30% 025K50K75K100K125K Tokens used, lower is better Pass rate GPT-5.2 (none): Pass rate 0.54%, tokens used 1,060 GPT-5.2 (low): Pass rate 1.09%, tokens used 6,955 GPT-5.2 (medium): Pass rate 2.43%, tokens used 18K GPT-5.2 (high): Pass rate 3.48%, tokens used 23K GPT-5.2 (xhigh): Pass rate 4.88%, tokens used 52K GPT-5.4 (none): Pass rate 0.85%, tokens used 2,101 GPT-5.4 (low): Pass rate 3.04%, tokens used 8,230 GPT-5.4 (medium): Pass rate 4.96%, tokens used 19K GPT-5.4 (high): Pass rate 6.99%, tokens used 27K GPT-5.4 (xhigh): Pass rate 8.87%, tokens used 45K GPT-5.5 (none): Pass rate 0.78%, tokens used 1,138 GPT-5.5 (low): Pass rate 2.41%, tokens used 2,768 GPT-5.5 (medium): Pass rate 5.89%, tokens used 11K GPT-5.5 (high): Pass rate 9.28%, tokens used 20K GPT-5.5 (xhigh): Pass rate 12%, tokens used 29K GPT-5.6 Luna (none): Pass rate 0.78%, tokens used 975 GPT-5.6 Luna (low): Pass rate 2.33%, tokens used 3,574 GPT-5.6 Luna (medium): Pass rate 4.7%, tokens used 16K GPT-5.6 Luna (high): Pass rate 8.04%, tokens used 32K GPT-5.6 Luna (xhigh): Pass rate 10.8%, tokens used 53K GPT-5.6 Luna (max): Pass rate 16.5%, tokens used 118K GPT-5.6 Terra (medium): Pass rate 13.6%, tokens used 16K GPT-5.6 Terra (high): Pass rate 16.2%, tokens used 22K GPT-5.6 Terra (xhigh): Pass rate 18.8%, tokens used 31K GPT-5.6 Terra (max): Pass rate 23.3%, tokens used 54K GPT-5.6 Sol (none): Pass rate 3.72%, tokens used 1,400 GPT-5.6 Sol (low): Pass rate 14.4%, tokens used 5,600 GPT-5.6 Sol (medium): Pass rate 22.5%, tokens used 14K GPT-5.6 Sol (high): Pass rate 24.4%, tokens used 20K GPT-5.6 Sol (xhigh): Pass rate 26.9%, tokens used 26K GPT-5.6 Sol (max): Pass rate 28.7%, tokens used 33K GPT-5.6 Terra (none): Pass rate 1.01%, tokens used 930 GPT-5.6 Terra (low): Pass rate 6.53%, tokens used 5,509 GPT-5.6 Sol (max)GPT-5.6 Sol (xhigh)GPT-5.6 Sol (high)GPT-5.6 Sol (medium)GPT-5.6 Sol (low)GPT-5.6 Terra (low)GPT-5.6 Sol (none)GPT-5.6 Terra (none) 0% 10% 20% 30% 025K50K75K100K125K Tokens used, lower is better Pass rate GPT-5.2 (none): Pass rate 0.54%, tokens used 1,060 GPT-5.2 (low): Pass rate 1.09%, tokens used 6,955 GPT-5.2 (medium): Pass rate 2.43%, tokens used 18K GPT-5.2 (high): Pass rate 3.48%, tokens used 23K GPT-5.2 (xhigh): Pass rate 4.88%, tokens used 52K GPT-5.4 (none): Pass rate 0.85%, tokens used 2,101 GPT-5.4 (low): Pass rate 3.04%, tokens used 8,230 GPT-5.4 (medium): Pass rate 4.96%, tokens used 19K GPT-5.4 (high): Pass rate 6.99%, tokens used 27K GPT-5.4 (xhigh): Pass rate 8.87%, tokens used 45K GPT-5.5 (none): Pass rate 0.78%, tokens used 1,138 GPT-5.5 (low): Pass rate 2.41%, tokens used 2,768 GPT-5.5 (medium): Pass rate 5.89%, tokens used 11K GPT-5.5 (high): Pass rate 9.28%, tokens used 20K GPT-5.5 (xhigh): Pass rate 12%, tokens used 29K GPT-5.6 Luna (none): Pass rate 0.78%, tokens used 975 GPT-5.6 Luna (low): Pass rate 2.33%, tokens used 3,574 GPT-5.6 Luna (medium): Pass rate 4.7%, tokens used 16K GPT-5.6 Luna (high): Pass rate 8.04%, tokens used 32K GPT-5.6 Luna (xhigh): Pass rate 10.8%, tokens used 53K GPT-5.6 Luna (max): Pass rate 16.5%, tokens used 118K GPT-5.6 Terra (medium): Pass rate 13.6%, tokens used 16K GPT-5.6 Terra (high): Pass rate 16.2%, tokens used 22K GPT-5.6 Terra (xhigh): Pass rate 18.8%, tokens used 31K GPT-5.6 Terra (max): Pass rate 23.3%, tokens used 54K GPT-5.6 Sol (none): Pass rate 3.72%, tokens used 1,400 GPT-5.6 Sol (low): Pass rate 14.4%, tokens used 5,600 GPT-5.6 Sol (medium): Pass rate 22.5%, tokens used 14K GPT-5.6 Sol (high): Pass rate 24.4%, tokens used 20K GPT-5.6 Sol (xhigh): Pass rate 26.9%, tokens used 26K GPT-5.6 Sol (max): Pass rate 28.7%, tokens used 33K GPT-5.6 Terra (none): Pass rate 1.01%, tokens used 930 GPT-5.6 Terra (low): Pass rate 6.53%, tokens used 5,509 GPT-5.6 Sol (max)GPT-5.6 Sol (xhigh)GPT-5.6 Sol (high)GPT-5.6 Sol (medium)GPT-5.6 Sol (low)GPT-5.6 Terra (low)GPT-5.6 Sol (none)GPT-5.6 Terra (none)
ModelSettingPass rateTokens used
GPT-5.6 Sol max 28.7% 33,168
GPT-5.6 Sol xhigh 26.9% 25,691
GPT-5.6 Sol high 24.4% 19,535
GPT-5.6 Terra max 23.3% 54,267
GPT-5.6 Sol medium 22.5% 14,440
GPT-5.6 Terra xhigh 18.8% 31,054
GPT-5.6 Luna max 16.5% 118,180
GPT-5.6 Terra high 16.2% 22,175
GPT-5.6 Sol low 14.4% 5,600
GPT-5.6 Terra medium 13.6% 15,935
GPT-5.5 xhigh 12% 28,721
GPT-5.6 Luna xhigh 10.8% 53,053
GPT-5.5 high 9.3% 19,822
GPT-5.4 xhigh 8.9% 44,661
GPT-5.6 Luna high 8% 32,293
GPT-5.4 high 7% 27,262
GPT-5.6 Terra low 6.5% 5,509
GPT-5.5 medium 5.9% 10,575
GPT-5.4 medium 5% 18,590
GPT-5.2 xhigh 4.9% 51,983
GPT-5.6 Luna medium 4.7% 15,579
GPT-5.6 Sol none 3.7% 1,400
GPT-5.2 high 3.5% 23,158
GPT-5.4 low 3% 8,230
GPT-5.2 medium 2.4% 18,137
GPT-5.5 low 2.4% 2,768
GPT-5.6 Luna low 2.3% 3,574
GPT-5.2 low 1.1% 6,955
GPT-5.6 Terra none 1% 929.9
GPT-5.4 none 0.9% 2,101
GPT-5.6 Luna none 0.8% 975.04
GPT-5.5 none 0.8% 1,138
GPT-5.2 none 0.5% 1,060

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
GPT-5.6 Sol
pro none low medium xhigh
31.5%
GPT-5.6 Terra
pro none low medium xhigh max
28.5%
GPT-5.6 Luna
pro none medium high max
23.6%
GPT-5.5
pro none medium high
20.5%
GPT-5.4
pro none medium xhigh
16.3%
GPT-5.2
pro none high
8.53%

Source: Official GeneBench Pro leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
GPT-5.6 Sol pro 31.5%
GPT-5.6 Sol max 28.7%
GPT-5.6 Terra pro 28.5%
GPT-5.6 Sol xhigh 26.9%
GPT-5.6 Sol high 24.4%
GPT-5.6 Luna pro 23.6%
GPT-5.6 Terra max 23.3%
GPT-5.6 Sol medium 22.5%
GPT-5.5 pro 20.5%
GPT-5.6 Terra xhigh 18.8%
GPT-5.6 Luna max 16.5%
GPT-5.4 pro 16.3%
GPT-5.6 Terra high 16.2%
Claude Opus 4.8 max available reasoning 16%
GPT-5.6 Sol low 14.4%
GPT-5.6 Terra medium 13.6%
GPT-5.5 xhigh 12%
GPT-5.6 Luna xhigh 10.8%
GPT-5.5 high 9.28%
GPT-5.4 xhigh 8.87%
GPT-5.2 pro 8.53%
Gemini 3.5 Flash max available reasoning 8.14%
GPT-5.6 Luna high 8.04%
GPT-5.4 high 6.99%
GPT-5.6 Terra low 6.53%
GPT-5.5 medium 5.89%
GPT-5.4 medium 4.96%
GPT-5.2 xhigh 4.88%
GPT-5.6 Luna medium 4.70%
GLM-5.2 max available reasoning 4.64%
GPT-5.6 Sol none 3.72%
GPT-5.2 high 3.48%
Gemini 3.1 Pro max available reasoning 3.10%
GPT-5.4 low 3.04%
GPT-5.2 medium 2.43%
DeepSeek-V4-Pro max available reasoning 2.42%
GPT-5.5 low 2.41%
GPT-5.6 Luna low 2.33%
Grok 4.3 max available reasoning 1.47%
GPT-5.2 low 1.09%
GPT-5.6 Terra none 1.01%
GPT-5.4 none 0.85%
GPT-5.5 none 0.78%
GPT-5.6 Luna none 0.78%
GPT-5.2 none 0.54%

2 models have held number one on the official leaderboard since its release in June 2026 4 models have held number one with claims included since its release in June 2026

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.

OpenAI
Own leaderboard Not tested since Jul 2026
GPT-5.6 Sol 31.5%
Lab claims 3 leaders
GPT-Rosalind-6 37.7%

Sources: Official GeneBench Pro leaderboard and labs’ model cards

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%) GPT-Rosalind-6 GPT-5.6 Sol GPT-5.5 GPT-5.4 GPT-5.2 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-Rosalind-6 GPT-5.5

Sources: Official GeneBench Pro leaderboard and labs’ model cards

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

OpenAI
DateLaunch postScoreAs the post writes it
Sep 3, 2026 GPT-6 Astra 37.1 GeneBench Pro
Jul 9, 2026 GPT-5.6 Luna 10.8 GeneBench Pro
Jul 9, 2026 GPT-5.6 Terra 23.3 GeneBench Pro
Jul 9, 2026 GPT-5.6 Sol 28.7 GeneBench Pro (Sol, max)

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
Oct 30, 2026 GPT-Rosalind-6 system card 37.7 GeneBench Pro v13, highest reasoning budget, single-attempt success Capabilities

The lab's number comes out ahead in 0 of 3 pairs that differ

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. Every benchmark.

Model Lab says Board says Gap
GPT-5.6 Luna 10.8% GeneBench Pro 23.6% Leaderboard, pro −13 pts
GPT-5.6 Terra 23.3% GeneBench Pro 28.5% Leaderboard, pro −5.2 pts
GPT-5.6 Sol 28.7% GeneBench Pro (Sol, max) 31.5% Leaderboard, pro −2.8 pts

More in Biology and chemistry

All 109 in Biology and chemistry
  • 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%
  • TOMG-Bench In use Models generate molecules from natural language requests in three tasks: molecule editing, molecule optimization and customized generation. Each output is checked for validity and for meeting the requested properties. GPT-3.5 Turbo 18.6%