LiveBench Theory of Mind: leaderboard
A LiveBench reasoning task added in November 2025 in place of the web of lies task, asking the model to reason about the internal states of other people in the situations it is given. Like the rest of LiveBench it is scored automatically against ground-truth answers rather than by a judge model or human voters, and the board shows the task average where higher is better.
Metric: Score. Source: livebench.ai. Saturation forecast: Around January 2027. 71 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.3 | 88.46 | #39 |
| 2 | GPT-5.4 (xHigh) | 88.46 | #79 (GPT-5.4) |
| 3 | Grok 4.6 | 86.54 | #26 |
| 4 | GPT-5.6 Terra (Max) | 86.54 | #41 (GPT-5.6 Terra) |
| 5 | GPT-6.1 Sol (Max) | 86.54 | #7 (GPT-6.1 Sol) |
| 6 | DeepSeek V4 Flash (0731) | 86.54 | #78 |
| 7 | Claude Haiku 5.5 (Max) | 86.54 | #40 (Claude Haiku 5.5) |
| 8 | Muse Wonder (Max) | 86.54 | #15 (Muse Wonder) |
| 9 | GPT-6.1 Sol (xHigh) | 86.54 | #7 (GPT-6.1 Sol) |
| 10 | Claude Sonnet 5.5 (Max) | 86.54 | #11 (Claude Sonnet 5.5) |
| 11 | GPT-5.5 (xHigh) | 84.61 | #31 (GPT-5.5) |
| 12 | GPT-5.6 Sol (Max) | 84.61 | #19 (GPT-5.6 Sol) |
| 13 | GPT-6 Astra (Max) | 84.61 | #6 (GPT-6 Astra) |
| 14 | Grok 4.7 (xHigh) | 84.61 | #25 (Grok 4.7) |
| 15 | GPT-6 Sol (Max) | 84.61 | #14 (GPT-6 Sol) |
No result here: #8 Gemini 4 Argon, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=livebench-theory-of-mind · How It Works · Data refreshed daily, snapshot 2026-10-20.