MysteryMechanism leaderboard
As of 6 Nov 2026, the top score on MysteryMechanism is 53.1% by GPT-6 Astra (OpenAI), ahead of Claude Opus 5.5 at 49.5%, across 29 model configurations with a published result (Vals AI, result dated 29 Oct 2026). The model can request observations of an unknown system and must infer the mechanism behind them, a test of active scientific reasoning. MysteryMechanism is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- GPT-6 Astra holds the top MysteryMechanism score at 53.1% (Vals AI, 29 Oct 2026).
- Claude Opus 5.5 is second on MysteryMechanism at 49.5% (Vals AI, 29 Oct 2026).
- MysteryMechanism: Working out a hidden scientific mechanism from sparse observations. Run by Vals AI (Vals AI).
- 29 model configurations have a published MysteryMechanism result on BenchLeader (BenchLeader, 6 Nov 2026).
- MysteryMechanism does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Working out a hidden scientific mechanism from sparse observations. Run by Vals AI.
- Published by
- Vals AI
- Category
- Reasoning
- Index weight
- Reference only
- Models
- 29
- Data as of
- 6 Nov 2026
Vals AI (vals.ai).
What the test looks like
The model can request observations of an unknown system and must infer the mechanism behind them, a test of active scientific reasoning.
How it is scored
Percent of mechanisms identified, run by Vals AI.
What to keep in mind
Small and unusual; a good complement to static exams.
- 1GPT-6 Astra (max)53.1%
- 2Claude Opus 5.549.5%
- 3Claude Sonnet 5.549.1%
- 4Claude Fable 5.147.8%
- 5GPT-6.1 Sol (max)46.4%
- 6Gemini 4 Argon (high)45.5%
- 7Claude Opus 537.4%
- 8Gemini 3.8 Flash (high)36.5%
- 9Muse Spark 1.3 (max)36.0%
- 10Muse Wonder Full (max)35.1%
- 11Claude Haiku 5.535.1%
- 12GPT-5.6 Sol (max)33.3%
- 13Grok 4.6 (high)30.6%
- 14GPT-6 Sol (max)30.2%
- 15Muse Wonder (max)27.5%
| # | |||||
|---|---|---|---|---|---|
| 1 |  GPT-6 Astramax | 53.1% | 67.8 | $1.56 | 2026-10-29 |
| 2 |  Claude Opus 5.5 | 49.5% | 63.4 | $4.62 | 2026-10-29 |
| 3 | Claude Sonnet 5.5 | 49.1% | – | $3.45 | 2026-10-29 |
| 4 | Claude Fable 5.1 | 47.8% | 66.4 | $5.63 | 2026-10-29 |
| 5 | GPT-6.1 Solmax | 46.4% | 67.3 | $0.285 | 2026-10-29 |
| 6 |  Gemini 4 Argonhigh | 45.5% | 66.0 | $4.59 | 2026-10-29 |
| 7 | Claude Opus 5 | 37.4% | 65.2 | $3.28 | 2026-10-29 |
| 8 | Gemini 3.8 Flashhigh | 36.5% | 62.0 | $1.73 | 2026-10-29 |
| 9 |  Muse Spark 1.3max | 36.0% | 62.0 | $0.852 | 2026-10-29 |
| 10 | Muse Wonder Fullmax | 35.1% | 62.9 | $4.16 | 2026-10-29 |
| 11 | Claude Haiku 5.5 | 35.1% | – | $0.541 | 2026-10-29 |
| 12 | GPT-5.6 Solmax | 33.3% | 65.4 | $0.963 | 2026-10-29 |
| 13 |  Grok 4.6high | 30.6% | 60.5 | $1.31 | 2026-10-29 |
| 14 | GPT-6 Solmax | 30.2% | 63.3 | $0.461 | 2026-10-29 |
| 15 | Muse Wondermax | 27.5% | 62.3 | $3.39 | 2026-10-29 |
| 16 |  Kimi K3.1 Promaxopen ↗ | 26.1% | – | $0.622 | 2026-10-29 |
| 17 | Grok 4.7xhigh | 25.2% | 58.3 | $1.84 | 2026-10-29 |
| 18 |  Qwen3.8 Maxmax | 23.9% | 58.8 | $1.31 | 2026-10-29 |
| 19 |  GLM 5.3maxopen ↗ | 23.0% | 59.0 | $1.53 | 2026-10-29 |
| 20 | Ember-1 | 22.5% | – | $1.29 | 2026-10-29 |
| 21 |  Beam 1.0open | 22.1% | 45.8 | – | 2026-10-29 |
| 22 |  MiMo-V2.6-Flashopen ↗ | 21.6% | 55.0 | $0.057 | 2026-10-29 |
| 23 |  DeepSeek V4.1 Flashhighopen ↗ | 21.2% | 55.7 | $0.164 | 2026-10-29 |
| 24 | Gemini 3.1 Prohigh | 20.7% | 60.1 | $1.07 | 2026-10-29 |
| 25 | GPT-6 Lunamax | 19.4% | 56.7 | $0.026 | 2026-10-29 |
| 26 |  Step 5 Previewopen | 18.5% | 54.9 | $0.597 | 2026-10-29 |
| 27 | MiMo-V2.6-Proopen ↗ | 15.3% | 57.3 | $0.154 | 2026-10-29 |
| 28 | GPT-5.6 Lunamax | 14.4% | 56.3 | $0.109 | 2026-10-29 |
| 29 |  Mistral Large 4highopen ↗ | 12.2% | 49.5 | $1.06 | 2026-10-29 |
More reasoning benchmarks · All LLM benchmarks
Related benchmarks
Top head-to-heads on MysteryMechanism
Cite as: BenchLeader, “MysteryMechanism leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/vals_mystery_mechanism.
MysteryMechanism: questions
- What does MysteryMechanism measure?
- The model can request observations of an unknown system and must infer the mechanism behind them, a test of active scientific reasoning. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on MysteryMechanism?
- As of 6 Nov 2026, the highest score on MysteryMechanism is 53.1% by GPT-6 Astra (result dated 29 Oct 2026), ahead of Claude Opus 5.5 at 49.5%. Scores come from Vals AI and are re-read daily.
- Which AI model leads MysteryMechanism?
- GPT-6 Astra leads MysteryMechanism with 53.1% as of 6 Nov 2026 (result dated 29 Oct 2026), ahead of Claude Opus 5.5 at 49.5%.
- How many models have MysteryMechanism results?
- 29 model configurations have a MysteryMechanism result on BenchLeader, all taken from Vals AI.
- Who runs MysteryMechanism and how often is it updated?
- MysteryMechanism is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does MysteryMechanism count toward the BenchLeader Index?
- No. MysteryMechanism is shown for reference but left out of the composite index.