Benchmark Almanac

DualChem

GPT-5.6 Sol 67.3% leads since October 2026
Not in model cards

Models answer chemistry questions written by experts. The researchers report answer accuracy and assess whether models recognize possible misuse.

Scale AI developed the 600-question benchmark under its NIST research agreement and evaluated more than 30 models. Higher refers to answer accuracy.

Released
October 2026
Kind
Dangerous capabilities
Level
University
Scored
Checked automatically
Answers
Multiple choice
Measures
Capability
Risk
CBRN
Language
English
In and out
Text → Text
Interaction
Single turn
Items from
Written by experts
Refresh
One fixed set
Run by
A host
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 21st, 2026
33
Entries in its paper, 32 models we track

GPT-5.6 Sol leads with 67.3% GPT-5.6 Sol leads with 67.3%

Best score per entry from the benchmark's paper, read on November 21st, 2026. A paper’s table doesn’t add newer models. 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. This view shows only models we track, each at its best.

  1. 1 GPT-5.6 Sol 67.3%
  2. 2 GPT-6 Astra 62.7%
  3. 3 GPT-6.1 Sol 61.7%
  4. 4 GPT-5.6 Terra 60.3%
  5. 5 Muse Wonder 60%
  6. 6 GPT-5.2 59.8%
  7. 7 Muse Spark 1.3 59.8%
  8. 8 GPT-5.6 Luna 59.2%
  9. 9 Claude Opus 4.8 58.3%
  10. 10 Claude Sonnet 5.5 58%
  11. 11 o3 57.5%
  12. 12 Muse Spark Open 56.8%
  13. 13 Claude Sonnet 5 56.7%
  14. 14 Gemini 4 Argon 55.2%
  15. 15 Claude Opus 4.5 53.3%
  16. 16 Gemini 3 Flash 52.5%
  17. 17 Gemini 2.5 Pro Jun 2025 52.3%
  18. 18 DeepSeek-V4-Pro 52.3%
  19. 19 o4-mini 52.2%
  20. 20 Claude Opus 5.5 50.2%
  21. 21 Claude Sonnet 4.5 49.8%
  22. 22 Claude Opus 4.1 48.8%
  23. 23 Gemini 3.5 Flash 48.7%
  24. 24 Gemini 3.1 Pro 45.7%
  25. 25 Llama 3.3 70B 45.2%
  26. 26 o3-mini 45%
  27. 27 DeepSeek-V3 Mar 2025 44.5%
  28. 28 Gemini 3 Pro 43.3%
  29. 29 Claude Fable 5 43.3%
  30. 30 Claude Fable 5.1 42.8%
  31. 31 Qwen3 235B 41.2%
  32. 32 Qwen3-32B 40.2%
  33. 33 Llama 3.1-8B 23.7%

Source: The DualChem paper

#EntryScoreSettingDateSource
1 GPT-5.6 Sol 67.3% Jul 2026 Official leaderboard
2 GPT-6 Astra 62.7% Sep 2026 Official leaderboard
3 GPT-6.1 Sol 61.7% Sep 2026 Official leaderboard
4 GPT-5.6 Terra 60.3% Jul 2026 Official leaderboard
5 Muse Wonder 60% Oct 2026 Official leaderboard
6 GPT-5.2 59.8% Dec 2025 Official leaderboard
7 Muse Spark 1.3 59.8% Sep 2026 Official leaderboard
8 GPT-5.6 Luna 59.2% Jul 2026 Official leaderboard
9 Claude Opus 4.8 58.3% May 2026 Official leaderboard
10 Claude Sonnet 5.5 58% Sep 2026 Official leaderboard
11 o3 57.5% Apr 2025 Official leaderboard
12 Muse Spark Open 56.8% Oct 2026 Official leaderboard
13 Claude Sonnet 5 56.7% Jun 2026 Official leaderboard
14 Gemini 4 Argon 55.2% Sep 2026 Official leaderboard
15 Claude Opus 4.5 53.3% Nov 2025 Official leaderboard
16 Gemini 3 Flash 52.5% Dec 2025 Official leaderboard
17 Gemini 2.5 Pro Jun 2025 52.3% Jun 2025 Official leaderboard
18 DeepSeek-V4-Pro 52.3% Apr 2026 Official leaderboard
19 o4-mini 52.2% Apr 2025 Official leaderboard
20 Claude Opus 5.5 50.2% Sep 2026 Official leaderboard
21 Claude Sonnet 4.5 49.8% Sep 2025 Official leaderboard
22 Claude Opus 4.1 48.8% Aug 2025 Official leaderboard
23 Gemini 3.5 Flash 48.7% May 2026 Official leaderboard
24 Gemini 3.1 Pro 45.7% Feb 2026 Official leaderboard
25 Llama 3.3 70B 45.2% Dec 2024 Official leaderboard
26 o3-mini 45% Jan 2025 Official leaderboard
27 DeepSeek-V3 Mar 2025 44.5% Mar 2025 Official leaderboard
28 Gemini 3 Pro 43.3% Nov 2025 Official leaderboard
29 Claude Fable 5 43.3% Jun 2026 Official leaderboard
30 Claude Fable 5.1 42.8% Sep 2026 Official leaderboard
31 Qwen3 235B 41.2% Official leaderboard
32 Qwen3-32B 40.2% Apr 2025 Official leaderboard
33 Llama 3.1-8B 23.7% Jul 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-5.6 Sol 67.3% Jul 2026 Official leaderboard
2 GPT-6 Astra 62.7% Sep 2026 Official leaderboard
3 GPT-6.1 Sol 61.7% Sep 2026 Official leaderboard
4 GPT-5.6 Terra 60.3% Jul 2026 Official leaderboard
5 Muse Wonder 60% Oct 2026 Official leaderboard
6 GPT-5.2 59.8% Dec 2025 Official leaderboard
7 Muse Spark 1.3 59.8% Sep 2026 Official leaderboard
8 GPT-5.6 Luna 59.2% Jul 2026 Official leaderboard
9 Claude Opus 4.8 58.3% May 2026 Official leaderboard
10 Claude Sonnet 5.5 58% Sep 2026 Official leaderboard
11 o3 57.5% Apr 2025 Official leaderboard
12 Muse Spark Open 56.8% Oct 2026 Official leaderboard
13 Claude Sonnet 5 56.7% Jun 2026 Official leaderboard
14 Gemini 4 Argon 55.2% Sep 2026 Official leaderboard
15 Claude Opus 4.5 53.3% Nov 2025 Official leaderboard
16 Gemini 3 Flash 52.5% Dec 2025 Official leaderboard
17 Gemini 2.5 Pro Jun 2025 52.3% Jun 2025 Official leaderboard
18 DeepSeek-V4-Pro 52.3% Apr 2026 Official leaderboard
19 o4-mini 52.2% Apr 2025 Official leaderboard
20 Claude Opus 5.5 50.2% Sep 2026 Official leaderboard
21 Claude Sonnet 4.5 49.8% Sep 2025 Official leaderboard
22 Claude Opus 4.1 48.8% Aug 2025 Official leaderboard
23 Gemini 3.5 Flash 48.7% May 2026 Official leaderboard
24 Gemini 3.1 Pro 45.7% Feb 2026 Official leaderboard
25 Llama 3.3 70B 45.2% Dec 2024 Official leaderboard
26 o3-mini 45% Jan 2025 Official leaderboard
27 DeepSeek-V3 Mar 2025 44.5% Mar 2025 Official leaderboard
28 Gemini 3 Pro 43.3% Nov 2025 Official leaderboard
29 Claude Fable 5 43.3% Jun 2026 Official leaderboard
30 Claude Fable 5.1 42.8% Sep 2026 Official leaderboard
31 Qwen3-32B 40.2% Apr 2025 Official leaderboard
32 Llama 3.1-8B 23.7% Jul 2024 Official leaderboard

Number one over time

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

OpenAI
Benchmark paper 1 leader
GPT-5.6 Sol 67.3%

Source: The DualChem paper

Scores over time

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

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.6 Sol GPT-5.2 o3 Llama 3.3 70B Llama 3.1-8B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) GPT-5.6 Sol GPT-5.2

Source: Official DualChem leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Safety and alignment

All 636 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%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

10 of 11 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the benchmark’s paper (33 entries).
  • Pass Models scored 32 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from October 2026, for Muse Spark Open.
  • Pass Frontier models 6 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 67.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on scale.com: 32 rows for models we track and 1 other entry at the last read.
  • Pass Last read November 21st, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check