Benchmark Almanac

BioTIER Permit

GPT-5.5 Rosalind 100% tops its leaderboard GPT-4.1 99.9% leads the models we track
Only in system cards

Models receive 144 benign biology prompts. The score measures the percentage correctly answered.

One track of BioTIER. The BioTIER record keeps lab scores that combine tracks.

Made by
SecureBio
Released
September 2026
Kind
Refusing harmful requests
Level
Expert
Scored
Graded by a model
Answers
Open-ended
Tasks
Kept private
Strongest model tested
GPT-6 Astra #3 on the capability index
Last in a system card
October 2026
Last model added
October 2026
Scores refreshed
November 12th, 2026
77
Entries on its own leaderboard, 69 models we track
1
System card reports it, from 1 lab

GPT-5.5 Rosalind tops the leaderboard with 100%

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is GPT-4.1, at 99.9%.

77 entries. Scroll sideways for the rest.

  1. 100.0 GPT-5.5 Rosalind
  2. 99.9 Gemini 3.1 Pro (Feb 26' Preview)
  3. 99.9 GPT-4.1
  4. 99.9 Mistral Medium 3.1
  5. 99.9 Kimi K2 Thinking
  6. 99.9 DeepSeek-V3.2
  7. 99.9 Mistral Large 3
  8. 99.9 Gemini 3.5 Flash
  9. 99.9 Qwen3.7-Max
  10. 99.9 Inkling
  11. 99.9 Qwen3.8-2.4T-A95B
  12. 99.9 Mistral Large 4
  13. 99.9 Nemotron 3.5 Super VL
  14. 99.9 Kimi K3.1 Pro
  15. 99.8 GPT-4.1 mini
  16. 99.8 Qwen3.5 397B-A17B
  17. 99.8 Inkling-Small
  18. 99.8 Gemini 3.6 Flash
  19. 99.8 Qwen 3.8 27B
  20. 99.7 Gemini 2.5 Flash (snapshot unspecified)
  21. 99.7 GPT-4o mini (snapshot unspecified)
  22. 99.7 GPT-4o
  23. 99.7 GPT-4o
  24. 99.7 Qwen3-235B-A22B-Thinking
  25. 99.7 GLM-5
  26. 99.7 Grok 4.20
  27. 99.7 GLM-5.2
  28. 99.7 Grok 4.5
  29. 99.7 Nemotron 3.5 Lightning 30B-A3B
  30. 99.7 DeepSeek V4 Pro 0813
  31. 99.7 DeepSeek V4.1 Flash 1017
  32. 99.6 Nemotron 3 Nano 30B-A3B
  33. 99.6 Kimi K2.6
  34. 99.6 GLM-5.3
  35. 99.5 Gemma 4 31B IT
  36. 99.5 Kimi K3
  37. 99.4 Llama 3.3 70B
  38. 99.4 Kimi K2.5
  39. 99.4 GLM-5.1
  40. 99.4 Gemini 3.8 Flash
  41. 99.3 GPT-5.6 Luna
  42. 99.3 GPT-5.6 Sol
  43. 99.3 GPT-5.6 Terra
  44. 99.3 Gemini 3.7 Flash
  45. 99.2 Nemotron 3 Super 120B A12B
  46. 99.2 gpt-oss-120b
  47. 99.2 GPT-5.2
  48. 99.2 GPT-5.5 Pro
  49. 99.2 Grok 4.6
  50. 99.2 Grok 4.7
  51. 99.1 Nemotron 3 Ultra 550B A55B
  52. 99.1 GPT-5.4
  53. 99.0 GPT-5.5
  54. 98.9 GPT-6 Luna
  55. 98.8 Muse Glimmer 30B
  56. 98.8 GPT-6.1 Sol
  57. 98.7 Gemini 3 Flash (Dec '25 Preview)
  58. 98.6 GPT-6.1 Terra
  59. 98.5 GPT-6 Astra
  60. 98.5 GPT-6 Sol
  61. 98.5 Beam 1.0
  62. 98.3 Claude Haiku 4.5
  63. 97.9 gpt-oss-20b
  64. 97.6 Gemini 4 Argon
  65. 95.3 DeepSeek-R1
  66. 94.3 Claude Opus 4.5
  67. 92.4 Muse Spark Open
  68. 86.5 Claude Opus 4.6
  69. 84.4 Claude Opus 4.8
  70. 82.8 Claude Opus 4.7
  71. 82.4 Claude Opus 5
  72. 81.6 Claude Sonnet 4.6
  73. 78.5 Claude Sonnet 5.5
  74. 78.3 Claude Sonnet 5
  75. 69.2 Claude Sonnet 4.5
  76. 67.8 Claude Haiku 5.5
  77. 56.7 Claude Opus 5.5
  1. 1 GPT-5.5 Rosalind xhigh 100%
  2. 2 Gemini 3.1 Pro (Feb 26' Preview) high 99.9%
  3. 3 GPT-4.1 99.9%
  4. 4 Mistral Medium 3.1 99.9%
  5. 5 Kimi K2 Thinking 99.9%
  6. 6 DeepSeek-V3.2 99.9%
  7. 7 Mistral Large 3 99.9%
  8. 8 Gemini 3.5 Flash high 99.9%
  9. 9 Qwen3.7-Max 99.9%
  10. 10 Inkling xhigh 99.9%
  11. 11 Qwen3.8-2.4T-A95B xhigh 99.9%
  12. 12 Mistral Large 4 high 99.9%
  13. 13 Nemotron 3.5 Super VL 99.9%
  14. 14 Kimi K3.1 Pro max 99.9%
  15. 15 GPT-4.1 mini 99.8%
  16. 16 Qwen3.5 397B-A17B 99.8%
  17. 17 Inkling-Small xhigh 99.8%
  18. 18 Gemini 3.6 Flash high 99.8%
  19. 19 Qwen 3.8 27B xhigh 99.8%
  20. 20 Gemini 2.5 Flash (snapshot unspecified) 16k 99.7%
  21. 21 GPT-4o mini (snapshot unspecified) 99.7%
  22. 22 GPT-4o Aug 2024 99.7%
  23. 23 GPT-4o Nov 2024 99.7%
  24. 24 Qwen3-235B-A22B-Thinking Jul 2025 99.7%
  25. 25 GLM-5 99.7%
  26. 26 Grok 4.20 reasoning 99.7%
  27. 27 GLM-5.2 max 99.7%
  28. 28 Grok 4.5 high 99.7%
  29. 29 Nemotron 3.5 Lightning 30B-A3B 99.7%
  30. 30 DeepSeek V4 Pro 0813 max 99.7%
  31. 31 DeepSeek V4.1 Flash 1017 max 99.7%
  32. 32 Nemotron 3 Nano 30B-A3B 99.6%
  33. 33 Kimi K2.6 99.6%
  34. 34 GLM-5.3 max 99.6%
  35. 35 Gemma 4 31B IT 99.5%
  36. 36 Kimi K3 max 99.5%
  37. 37 Llama 3.3 70B 99.4%
  38. 38 Kimi K2.5 99.4%
  39. 39 GLM-5.1 99.4%
  40. 40 Gemini 3.8 Flash high 99.4%
  41. 41 GPT-5.6 Luna max 99.3%
  42. 42 GPT-5.6 Sol max 99.3%
  43. 43 GPT-5.6 Terra max 99.3%
  44. 44 Gemini 3.7 Flash high 99.3%
  45. 45 Nemotron 3 Super 120B A12B 99.2%
  46. 46 gpt-oss-120b 99.2%
  47. 47 GPT-5.2 xhigh 99.2%
  48. 48 GPT-5.5 Pro xhigh 99.2%
  49. 49 Grok 4.6 xhigh 99.2%
  50. 50 Grok 4.7 xhigh 99.2%
  51. 51 Nemotron 3 Ultra 550B A55B 99.1%
  52. 52 GPT-5.4 xhigh 99.1%
  53. 53 GPT-5.5 xhigh 99%
  54. 54 GPT-6 Luna max 98.9%
  55. 55 Muse Glimmer 30B xhigh 98.8%
  56. 56 GPT-6.1 Sol max 98.8%
  57. 57 Gemini 3 Flash (Dec '25 Preview) high 98.7%
  58. 58 GPT-6.1 Terra max 98.6%
  59. 59 GPT-6 Astra max 98.5%
  60. 60 GPT-6 Sol max 98.5%
  61. 61 Beam 1.0 max 98.5%
  62. 62 Claude Haiku 4.5 16k 98.3%
  63. 63 gpt-oss-20b 97.9%
  64. 64 Gemini 4 Argon high 97.6%
  65. 65 DeepSeek-R1 May 2025 95.3%
  66. 66 Claude Opus 4.5 16k 94.3%
  67. 67 Muse Spark Open xhigh 92.4%
  68. 68 Claude Opus 4.6 max 86.5%
  69. 69 Claude Opus 4.8 max 84.4%
  70. 70 Claude Opus 4.7 max 82.8%
  71. 71 Claude Opus 5 max 82.4%
  72. 72 Claude Sonnet 4.6 max 81.6%
  73. 73 Claude Sonnet 5.5 max 78.5%
  74. 74 Claude Sonnet 5 max 78.3%
  75. 75 Claude Sonnet 4.5 16k 69.2%
  76. 76 Claude Haiku 5.5 max 67.8%
  77. 77 Claude Opus 5.5 max 56.7%

Source: Official BioTIER Permit leaderboard

#EntryScoreSettingDateSource
1 GPT-5.5 Rosalind 100% xhigh Official leaderboard
2 Gemini 3.1 Pro (Feb 26' Preview) 99.9% high Official leaderboard
3 GPT-4.1 99.9% Apr 2025 Official leaderboard
4 Mistral Medium 3.1 99.9% Aug 2025 Official leaderboard
5 Kimi K2 Thinking 99.9% Nov 2025 Official leaderboard
6 DeepSeek-V3.2 99.9% Dec 2025 Official leaderboard
7 Mistral Large 3 99.9% Dec 2025 Official leaderboard
8 Gemini 3.5 Flash 99.9% high May 2026 Official leaderboard
9 Qwen3.7-Max 99.9% May 2026 Official leaderboard
10 Inkling 99.9% xhigh Jul 2026 Official leaderboard
11 Qwen3.8-2.4T-A95B 99.9% xhigh Aug 2026 Official leaderboard
12 Mistral Large 4 99.9% high Oct 2026 Official leaderboard
13 Nemotron 3.5 Super VL 99.9% Oct 2026 Official leaderboard
14 Kimi K3.1 Pro 99.9% max Oct 2026 Official leaderboard
15 GPT-4.1 mini 99.8% Apr 2025 Official leaderboard
16 Qwen3.5 397B-A17B 99.8% Feb 2026 Official leaderboard
17 Inkling-Small 99.8% xhigh Jul 2026 Official leaderboard
18 Gemini 3.6 Flash 99.8% high Jul 2026 Official leaderboard
19 Qwen 3.8 27B 99.8% xhigh Aug 2026 Official leaderboard
20 Gemini 2.5 Flash (snapshot unspecified) 99.7% 16k Official leaderboard
21 GPT-4o mini (snapshot unspecified) 99.7% Official leaderboard
22 GPT-4o Aug 2024 99.7% Aug 2024 Official leaderboard
23 GPT-4o Nov 2024 99.7% Nov 2024 Official leaderboard
24 Qwen3-235B-A22B-Thinking Jul 2025 99.7% Jul 2025 Official leaderboard
25 GLM-5 99.7% Feb 2026 Official leaderboard
26 Grok 4.20 99.7% reasoning Feb 2026 Official leaderboard
27 GLM-5.2 99.7% max Jun 2026 Official leaderboard
28 Grok 4.5 99.7% high Jul 2026 Official leaderboard
29 Nemotron 3.5 Lightning 30B-A3B 99.7% Aug 2026 Official leaderboard
30 DeepSeek V4 Pro 0813 99.7% max Aug 2026 Official leaderboard
31 DeepSeek V4.1 Flash 1017 99.7% max Oct 2026 Official leaderboard
32 Nemotron 3 Nano 30B-A3B 99.6% Dec 2025 Official leaderboard
33 Kimi K2.6 99.6% Apr 2026 Official leaderboard
34 GLM-5.3 99.6% max Aug 2026 Official leaderboard
35 Gemma 4 31B IT 99.5% Apr 2026 Official leaderboard
36 Kimi K3 99.5% max Jul 2026 Official leaderboard
37 Llama 3.3 70B 99.4% Dec 2024 Official leaderboard
38 Kimi K2.5 99.4% Jan 2026 Official leaderboard
39 GLM-5.1 99.4% Apr 2026 Official leaderboard
40 Gemini 3.8 Flash 99.4% high Sep 2026 Official leaderboard
41 GPT-5.6 Luna 99.3% max Jul 2026 Official leaderboard
42 GPT-5.6 Sol 99.3% max Jul 2026 Official leaderboard
43 GPT-5.6 Terra 99.3% max Jul 2026 Official leaderboard
44 Gemini 3.7 Flash 99.3% high Aug 2026 Official leaderboard
45 Nemotron 3 Super 120B A12B 99.2% Official leaderboard
46 gpt-oss-120b 99.2% Aug 2025 Official leaderboard
47 GPT-5.2 99.2% xhigh Dec 2025 Official leaderboard
48 GPT-5.5 Pro 99.2% xhigh Apr 2026 Official leaderboard
49 Grok 4.6 99.2% xhigh Aug 2026 Official leaderboard
50 Grok 4.7 99.2% xhigh Sep 2026 Official leaderboard
51 Nemotron 3 Ultra 550B A55B 99.1% Official leaderboard
52 GPT-5.4 99.1% xhigh Mar 2026 Official leaderboard
53 GPT-5.5 99% xhigh Apr 2026 Official leaderboard
54 GPT-6 Luna 98.9% max Sep 2026 Official leaderboard
55 Muse Glimmer 30B 98.8% xhigh Official leaderboard
56 GPT-6.1 Sol 98.8% max Sep 2026 Official leaderboard
57 Gemini 3 Flash (Dec '25 Preview) 98.7% high Official leaderboard
58 GPT-6.1 Terra 98.6% max Oct 2026 Official leaderboard
59 GPT-6 Astra 98.5% max Sep 2026 Official leaderboard
60 GPT-6 Sol 98.5% max Sep 2026 Official leaderboard
61 Beam 1.0 98.5% max Oct 2026 Official leaderboard
62 Claude Haiku 4.5 98.3% 16k Oct 2025 Official leaderboard
63 gpt-oss-20b 97.9% Aug 2025 Official leaderboard
64 Gemini 4 Argon 97.6% high Sep 2026 Official leaderboard
65 DeepSeek-R1 May 2025 95.3% May 2025 Official leaderboard
66 Claude Opus 4.5 94.3% 16k Nov 2025 Official leaderboard
67 Muse Spark Open 92.4% xhigh Oct 2026 Official leaderboard
68 Claude Opus 4.6 86.5% max Feb 2026 Official leaderboard
69 Claude Opus 4.8 84.4% max May 2026 Official leaderboard
70 Claude Opus 4.7 82.8% max Apr 2026 Official leaderboard
71 Claude Opus 5 82.4% max Jul 2026 Official leaderboard
72 Claude Sonnet 4.6 81.6% max Feb 2026 Official leaderboard
73 Claude Sonnet 5.5 78.5% max Sep 2026 Official leaderboard
74 Claude Sonnet 5 78.3% max Jun 2026 Official leaderboard
75 Claude Sonnet 4.5 69.2% 16k Sep 2025 Official leaderboard
76 Claude Haiku 5.5 67.8% max Oct 2026 Official leaderboard
77 Claude Opus 5.5 56.7% max Sep 2026 Official leaderboard

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

Number one over time

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

OpenAI
Own leaderboard Passed 85% in Sep 2026, by GPT-4.1
GPT-4.1 99.9%

Source: Official BioTIER Permit leaderboard

GPT-4o on the official leaderboard passed 85% in August 2024

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% 2025 2026 2027 Near the ceiling (85%) GPT-4.1 GPT-4o 0% 20% 40% 60% 80% 100% 2025 2026 2027 Near the ceiling (85%) GPT-4.1

Sources: Official BioTIER Permit 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
Oct 5, 2026 Beam 1.0 technical report 97.2 BioTIER permit split accuracy 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. Every benchmark.

Model Lab says Board says Gap
Beam 1.0 97.2% BioTIER permit split accuracy 98.5% Leaderboard, max −1.3 pts

More in Safety and alignment

All 465 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • DistressBench In use Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics. Muse Spark 1.2 88.4%
  • Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
  • IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%