Benchmark Almanac

Human Pathogen Capabilities Test v2 (HPCT v2)

Claude Opus 5 80% leads since Jul 2026 Not in model cards

Models answer multiple-response questions on a small set of weaponizable pathogens that biosecurity experts rank highest priority. A revised version of HPCT.

Shown as HPCT v2 beside HPCT on the SecureBio dashboard. Not public. Revised version so scores are not comparable with HPCT v1.

Made by SecureBio Released 2026 Hazardous knowledge Tasks kept private Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Research frontier Researchers, on open problems or tasks that take days
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Multiple choice
Leaderboard securebio.org
74
Entries on its own leaderboard, 71 models we track

Claude Opus 5 leads with 80%

Best score per entry from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

74 entries. Scroll sideways for the rest.

  1. 80.0 Claude Opus 5
  2. 75.2 Claude Opus 4.6
  3. 74.6 Gemini 4 Argon
  4. 71.5 Claude Opus 4.5
  5. 70.6 GPT-5.6 Sol
  6. 70.0 GPT-5.6 Terra
  7. 67.9 Claude Opus 4.7
  8. 66.3 Claude Opus 4.8
  9. 66.2 GPT-5.5
  10. 65.8 Gemini 3.6 Flash
  11. 65.4 Gemini 3.1 Pro
  12. 65.3 GPT-5.4
  13. 65.2 GPT-5.5 Pro
  14. 65.0 Claude Sonnet 5
  15. 64.7 Gemini 3.5 Flash
  16. 64.6 GPT-6 Astra
  17. 64.6 GPT-6 Sol
  18. 64.4 Gemini 3.8 Flash
  19. 64.0 GPT-5.2 Pro
  20. 63.9 GPT-5.4 Pro
  21. 63.6 Gemini 3.7 Flash
  22. 63.2 GPT-5.2
  23. 62.5 GPT-5.6 Luna
  24. 62.1 Grok 4.6
  25. 61.7 Kimi K3.1 Pro
  26. 61.2 Grok 4.5
  27. 60.2 Kimi K3
  28. 60.0 Claude Sonnet 4.6
  29. 59.6 GLM-5.2
  30. 59.6 Beam 1.0
  31. 57.9 Qwen3.8-2.4T-A95B
  32. 56.2 Grok 4.20
  33. 56.0 Gemini 3 Flash
  34. 54.5 Kimi K2.6
  35. 54.1 Grok 4.7
  36. 53.6 Grok 4.20 Multi-agent
  37. 53.2 GLM-5.3
  38. 52.7 DeepSeek-V4-Pro
  39. 51.8 Nemotron 3 Ultra 550B A55B
  40. 51.6 Grok-3 mini
  41. 50.9 Grok 4 Fast
  42. 50.9 Inkling
  43. 50.4 Kimi K2 Thinking
  44. 50.3 GLM-5.1
  45. 49.7 Grok 4
  46. 49.3 Grok 3
  47. 49.2 Grok 4.1 Fast
  48. 48.6 Qwen 3.8 27B
  49. 47.3 Qwen3.6 27B
  50. 46.4 DeepSeek-R1
  51. 44.6 Nemotron 3 Super 120B A12B
  52. 44.6 gpt-oss-120b
  53. 43.8 Gemma 4 31B IT
  54. 43.5 Gemini 2.5 Flash
  55. 43.4 Kimi K2.5
  56. 43.1 Qwen3-235B-A22B-Thinking
  57. 42.5 Claude Haiku 4.5
  58. 39.8 DeepSeek-V3.1
  59. 37.9 gpt-oss-20b
  60. 37.4 DeepSeek-V3.2
  61. 36.7 Nemotron 3 Nano 30B-A3B
  62. 34.2 GPT-4.1
  63. 33.4 Qwen3-Next-80B-A3B-Instruct
  64. 31.6 GPT-4o
  65. 31.5 GPT-4.1 mini
  66. 30.8 Mistral Large 3
  67. 30.0 GPT-4o
  68. 27.9 Mistral Small 3.2
  69. 27.5 Muse Glimmer 30B
  70. 27.5 Mistral Small 4
  71. 24.8 Llama 3.3 70B
  72. 24.7 Mistral Medium 3
  73. 24.3 Mistral Medium 3.1
  74. 13.1 GPT-4o mini
  1. 1 Claude Opus 5 max 80%
  2. 2 Claude Opus 4.6 max 75.2%
  3. 3 Gemini 4 Argon high 74.6%
  4. 4 Claude Opus 4.5 16k 71.5%
  5. 5 GPT-5.6 Sol max 70.6%
  6. 6 GPT-5.6 Terra max 70%
  7. 7 Claude Opus 4.7 max 67.9%
  8. 8 Claude Opus 4.8 max 66.3%
  9. 9 GPT-5.5 xhigh 66.2%
  10. 10 Gemini 3.6 Flash high 65.8%
  11. 11 Gemini 3.1 Pro high 65.4%
  12. 12 GPT-5.4 xhigh 65.3%
  13. 13 GPT-5.5 Pro xhigh 65.2%
  14. 14 Claude Sonnet 5 max 65%
  15. 15 Gemini 3.5 Flash high 64.7%
  16. 16 GPT-6 Astra max 64.6%
  17. 17 GPT-6 Sol max 64.6%
  18. 18 Gemini 3.8 Flash high 64.4%
  19. 19 GPT-5.2 Pro xhigh 64%
  20. 20 GPT-5.4 Pro xhigh 63.9%
  21. 21 Gemini 3.7 Flash high 63.6%
  22. 22 GPT-5.2 xhigh 63.2%
  23. 23 GPT-5.6 Luna max 62.5%
  24. 24 Grok 4.6 xhigh 62.1%
  25. 25 Kimi K3.1 Pro max 61.7%
  26. 26 Grok 4.5 high 61.2%
  27. 27 Kimi K3 max 60.2%
  28. 28 Claude Sonnet 4.6 max 60%
  29. 29 GLM-5.2 max 59.6%
  30. 30 Beam 1.0 max 59.6%
  31. 31 Qwen3.8-2.4T-A95B xhigh 57.9%
  32. 32 Grok 4.20 reasoning 56.2%
  33. 33 Gemini 3 Flash high 56%
  34. 34 Kimi K2.6 54.5%
  35. 35 Grok 4.7 xhigh 54.1%
  36. 36 Grok 4.20 Multi-agent 53.6%
  37. 37 GLM-5.3 max 53.2%
  38. 38 DeepSeek-V4-Pro max 52.7%
  39. 39 Nemotron 3 Ultra 550B A55B 51.8%
  40. 40 Grok-3 mini high 51.6%
  41. 41 Grok 4 Fast reasoning 50.9%
  42. 42 Inkling xhigh 50.9%
  43. 43 Kimi K2 Thinking 50.4%
  44. 44 GLM-5.1 50.3%
  45. 45 Grok 4 49.7%
  46. 46 Grok 3 49.3%
  47. 47 Grok 4.1 Fast reasoning 49.2%
  48. 48 Qwen 3.8 27B xhigh 48.6%
  49. 49 Qwen3.6 27B 47.3%
  50. 50 DeepSeek-R1 May 2025 46.4%
  51. 51 Nemotron 3 Super 120B A12B 44.6%
  52. 52 gpt-oss-120b 44.6%
  53. 53 Gemma 4 31B IT 43.8%
  54. 54 Gemini 2.5 Flash Jun 2025, 16k 43.5%
  55. 55 Kimi K2.5 43.4%
  56. 56 Qwen3-235B-A22B-Thinking Jul 2025 43.1%
  57. 57 Claude Haiku 4.5 16k 42.5%
  58. 58 DeepSeek-V3.1 39.8%
  59. 59 gpt-oss-20b 37.9%
  60. 60 DeepSeek-V3.2 37.4%
  61. 61 Nemotron 3 Nano 30B-A3B 36.7%
  62. 62 GPT-4.1 34.2%
  63. 63 Qwen3-Next-80B-A3B-Instruct 33.4%
  64. 64 GPT-4o Nov 2024 31.6%
  65. 65 GPT-4.1 mini 31.5%
  66. 66 Mistral Large 3 30.8%
  67. 67 GPT-4o Aug 2024 30%
  68. 68 Mistral Small 3.2 27.9%
  69. 69 Muse Glimmer 30B xhigh 27.5%
  70. 70 Mistral Small 4 27.5%
  71. 71 Llama 3.3 70B 24.8%
  72. 72 Mistral Medium 3 24.7%
  73. 73 Mistral Medium 3.1 24.3%
  74. 74 GPT-4o mini 13.1%
#EntryScoreSettingDateSource
1 Claude Opus 5 80% max Jul 2026 Official leaderboard
2 Claude Opus 4.6 75.2% max Feb 2026 Official leaderboard
3 Gemini 4 Argon 74.6% high Sep 2026 Official leaderboard
4 Claude Opus 4.5 71.5% 16k Nov 2025 Official leaderboard
5 GPT-5.6 Sol 70.6% max Jul 2026 Official leaderboard
6 GPT-5.6 Terra 70% max Jul 2026 Official leaderboard
7 Claude Opus 4.7 67.9% max Apr 2026 Official leaderboard
8 Claude Opus 4.8 66.3% max May 2026 Official leaderboard
9 GPT-5.5 66.2% xhigh Apr 2026 Official leaderboard
10 Gemini 3.6 Flash 65.8% high Jul 2026 Official leaderboard
11 Gemini 3.1 Pro 65.4% high Feb 2026 Official leaderboard
12 GPT-5.4 65.3% xhigh Mar 2026 Official leaderboard
13 GPT-5.5 Pro 65.2% xhigh Apr 2026 Official leaderboard
14 Claude Sonnet 5 65% max Jun 2026 Official leaderboard
15 Gemini 3.5 Flash 64.7% high May 2026 Official leaderboard
16 GPT-6 Astra 64.6% max Sep 2026 Official leaderboard
17 GPT-6 Sol 64.6% max Sep 2026 Official leaderboard
18 Gemini 3.8 Flash 64.4% high Sep 2026 Official leaderboard
19 GPT-5.2 Pro 64% xhigh Dec 2025 Official leaderboard
20 GPT-5.4 Pro 63.9% xhigh Mar 2026 Official leaderboard
21 Gemini 3.7 Flash 63.6% high Aug 2026 Official leaderboard
22 GPT-5.2 63.2% xhigh Dec 2025 Official leaderboard
23 GPT-5.6 Luna 62.5% max Jul 2026 Official leaderboard
24 Grok 4.6 62.1% xhigh Aug 2026 Official leaderboard
25 Kimi K3.1 Pro 61.7% max Oct 2026 Official leaderboard
26 Grok 4.5 61.2% high Jul 2026 Official leaderboard
27 Kimi K3 60.2% max Jul 2026 Official leaderboard
28 Claude Sonnet 4.6 60% max Feb 2026 Official leaderboard
29 GLM-5.2 59.6% max Jun 2026 Official leaderboard
30 Beam 1.0 59.6% max Oct 2026 Official leaderboard
31 Qwen3.8-2.4T-A95B 57.9% xhigh Aug 2026 Official leaderboard
32 Grok 4.20 56.2% reasoning Feb 2026 Official leaderboard
33 Gemini 3 Flash 56% high Dec 2025 Official leaderboard
34 Kimi K2.6 54.5% Apr 2026 Official leaderboard
35 Grok 4.7 54.1% xhigh Sep 2026 Official leaderboard
36 Grok 4.20 Multi-agent 53.6% Mar 2026 Official leaderboard
37 GLM-5.3 53.2% max Aug 2026 Official leaderboard
38 DeepSeek-V4-Pro 52.7% max Apr 2026 Official leaderboard
39 Nemotron 3 Ultra 550B A55B 51.8% Official leaderboard
40 Grok-3 mini 51.6% high Apr 2025 Official leaderboard
41 Grok 4 Fast 50.9% reasoning Sep 2025 Official leaderboard
42 Inkling 50.9% xhigh Jul 2026 Official leaderboard
43 Kimi K2 Thinking 50.4% Nov 2025 Official leaderboard
44 GLM-5.1 50.3% Apr 2026 Official leaderboard
45 Grok 4 49.7% Jul 2025 Official leaderboard
46 Grok 3 49.3% Apr 2025 Official leaderboard
47 Grok 4.1 Fast 49.2% reasoning Nov 2025 Official leaderboard
48 Qwen 3.8 27B 48.6% xhigh Aug 2026 Official leaderboard
49 Qwen3.6 27B 47.3% Apr 2026 Official leaderboard
50 DeepSeek-R1 May 2025 46.4% May 2025 Official leaderboard
51 Nemotron 3 Super 120B A12B 44.6% Official leaderboard
52 gpt-oss-120b 44.6% Aug 2025 Official leaderboard
53 Gemma 4 31B IT 43.8% Apr 2026 Official leaderboard
54 Gemini 2.5 Flash Jun 2025 43.5% 16k Jun 2025 Official leaderboard
55 Kimi K2.5 43.4% Jan 2026 Official leaderboard
56 Qwen3-235B-A22B-Thinking Jul 2025 43.1% Jul 2025 Official leaderboard
57 Claude Haiku 4.5 42.5% 16k Oct 2025 Official leaderboard
58 DeepSeek-V3.1 39.8% Aug 2025 Official leaderboard
59 gpt-oss-20b 37.9% Aug 2025 Official leaderboard
60 DeepSeek-V3.2 37.4% Dec 2025 Official leaderboard
61 Nemotron 3 Nano 30B-A3B 36.7% Dec 2025 Official leaderboard
62 GPT-4.1 34.2% Apr 2025 Official leaderboard
63 Qwen3-Next-80B-A3B-Instruct 33.4% Sep 2025 Official leaderboard
64 GPT-4o Nov 2024 31.6% Nov 2024 Official leaderboard
65 GPT-4.1 mini 31.5% Apr 2025 Official leaderboard
66 Mistral Large 3 30.8% Dec 2025 Official leaderboard
67 GPT-4o Aug 2024 30% Aug 2024 Official leaderboard
68 Mistral Small 3.2 27.9% Jun 2025 Official leaderboard
69 Muse Glimmer 30B 27.5% xhigh Official leaderboard
70 Mistral Small 4 27.5% Mar 2026 Official leaderboard
71 Llama 3.3 70B 24.8% Dec 2024 Official leaderboard
72 Mistral Medium 3 24.7% May 2025 Official leaderboard
73 Mistral Medium 3.1 24.3% Aug 2025 Official leaderboard
74 GPT-4o mini 13.1% Jul 2024 Official leaderboard

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
Grok 4.20
non-reasoni… reasoning
56.2%
Grok 4 Fast
reasoning
50.9%
Grok 4.1 Fast
reasoning
49.2%
ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5 max 80%
Claude Opus 4.6 max 75.2%
Gemini 4 Argon high 74.6%
Claude Opus 4.5 16k 71.5%
GPT-5.6 Sol max 70.6%
GPT-5.6 Terra max 70%
Claude Opus 4.7 max 67.9%
Claude Opus 4.8 max 66.3%
GPT-5.5 xhigh 66.2%
Gemini 3.6 Flash high 65.8%
Gemini 3.1 Pro high 65.4%
GPT-5.4 xhigh 65.3%
GPT-5.5 Pro xhigh 65.2%
Claude Sonnet 5 max 65%
Gemini 3.5 Flash high 64.7%
GPT-6 Astra max 64.6%
GPT-6 Sol max 64.6%
Gemini 3.8 Flash high 64.4%
GPT-5.2 Pro xhigh 64%
GPT-5.4 Pro xhigh 63.9%
Gemini 3.7 Flash high 63.6%
GPT-5.2 xhigh 63.2%
GPT-5.6 Luna max 62.5%
Grok 4.6 xhigh 62.1%
Kimi K3.1 Pro max 61.7%
Grok 4.5 high 61.2%
Kimi K3 max 60.2%
Claude Sonnet 4.6 max 60%
GLM-5.2 max 59.6%
Beam 1.0 max 59.6%
Qwen3.8-2.4T-A95B xhigh 57.9%
Grok 4.20 reasoning 56.2%
Gemini 3 Flash high 56%
Kimi K2.6 54.5%
Grok 4.7 xhigh 54.1%
Grok 4.20 Multi-agent 53.6%
GLM-5.3 max 53.2%
DeepSeek-V4-Pro max 52.7%
Grok-3 mini high 51.6%
Grok 4 Fast reasoning 50.9%
Inkling xhigh 50.9%
Kimi K2 Thinking 50.4%
GLM-5.1 50.3%
Grok 4 49.7%
Grok 3 49.3%
Grok 4.1 Fast reasoning 49.2%
Qwen 3.8 27B xhigh 48.6%
Qwen3.6 27B 47.3%
DeepSeek-R1 May 2025 46.4%
gpt-oss-120b 44.6%
Gemma 4 31B IT 43.8%
Gemini 2.5 Flash Jun 2025 16k 43.5%
Kimi K2.5 43.4%
Qwen3-235B-A22B-Thinking Jul 2025 43.1%
Claude Haiku 4.5 16k 42.5%
DeepSeek-V3.1 39.8%
gpt-oss-20b 37.9%
DeepSeek-V3.2 37.4%
Grok 4.20 non-reasoning 36.8%
Nemotron 3 Nano 30B-A3B 36.7%
GPT-4.1 34.2%
Grok 4 Fast 34.1%
Grok 4.1 Fast 34.1%
Qwen3-Next-80B-A3B-Instruct 33.4%
GPT-4o Nov 2024 31.6%
GPT-4.1 mini 31.5%
Mistral Large 3 30.8%
GPT-4o Aug 2024 30%
Mistral Small 3.2 27.9%
Mistral Small 4 27.5%
Llama 3.3 70B 24.8%
Mistral Medium 3 24.7%
Mistral Medium 3.1 24.3%
GPT-4o mini 13.1%

3 models have held number one since its release in Jan 2026

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

Anthropic
Own leaderboard 3 leaders
Claude Opus 5 80%

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% 2025 2026 2027 Near the ceiling (85%) Claude Opus 5 Claude Opus 4.6 Grok-3 mini GPT-4o GPT-4o mini 0% 20% 40% 60% 80% 100% 2025 2026 2027 Near the ceiling (85%) Claude Opus 5 Claude Opus 4.5

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

OpenAI Meta Anthropic
Human Pathogen Capabilities Test v2 (HPCT v2) 6 leaders
Claude Opus 5 80%
Human Pathogen Capabilities Test 3 leaders
GPT-5.6 Sol 68.4%
Human Pathogen Capabilities Test v2 (HPCT v2) 3 leaders
Claude Opus 5 80%
Replaces Human Pathogen Capabilities Test

More in Biology and chemistry

All 96 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%
  • 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-6 Astra 37.1%
  • 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-6 Astra 60.3%
  • MolecularIQ In use Models answer questions about molecular graphs, such as counting, indexing and constrained generation, and a symbolic checker verifies each answer. Intern-S2-397B 62.4%
  • 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. Intern-S2-397B 66.8%
  • BABE Fading Models reason from experimental results in peer-reviewed biology papers to draw conclusions, with causal and cross-scale inference. Seed2.0 Lite (0428) 57.9%