TestsBenchmarks · Reasoning
Humanity's Last Exam (with tools)
HLE with search/code tools (vendors blocklist HLE-discussing sources).
Measured by independent testersIndependent Reported by the makerVendor-reported
| Thinking levelSetting | |||||||
|---|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Anthropic | 67.7% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 22 Sep 2026 |
| 2 | Claude Fable 5.1 | Anthropic | 65.1% | Extra highadaptive thinking, effort=xhigh | Maker's own figureVendor-reportedAnthropic ↗ | — | 1 Sep 2026 |
| 3 | Claude Sonnet 5.5 | Anthropic | 64.5% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 28 Sep 2026 |
| 4 | DeepSeek V4.1 Flash | DeepSeek | 63.9% | Maxreasoning_effort=100 (max) | Maker's own figureVendor-reportedDeepSeek ↗ | — | 10 Sep 2026 |
| 5 | Claude Fable 5 | Anthropic | 63.8% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 1 Sep 2026 |
| 6 | Claude Opus 5 | Anthropic | 63.6% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 22 Sep 2026 |
| 7 | GLM-5.3 | Z.ai (Zhipu) | 62.5% | Maxreasoning_effort=max | Maker's own figureVendor-reportedZ.ai ↗ | — | 14 Aug 2026 |
| 8 | Muse Spark 1.1 | Meta (Meta Superintelligence Labs, Muse) | 62.1% | Defaulteffort not stated | Maker's own figureVendor-reportedMeta ↗ | — | 9 Jul 2026 |
| 9 | DeepSeek V4 Pro (0813) | DeepSeek | 60.0% | Maxreasoning_effort=max | Maker's own figureVendor-reportedDeepSeek ↗ | — | 13 Aug 2026 |
| 10 | Claude Opus 4.8 | Anthropic | 57.9% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 24 Jul 2026 |
| 11 | GPT-6 Astra | OpenAI | 57.2% | Defaultbest score across efforts (OpenAI table: "maximum at any effort") | Maker's own figureVendor-reportedOpenAI ↗ | — | 3 Sep 2026 |
| 12 | GPT-5.5 Pro | OpenAI | 57.2% | Extra highGPT-5.5 Pro, reasoning effort=xhigh | Maker's own figureVendor-reportedOpenAI ↗ | — | 23 Apr 2026 |
| 13 | Qwen3.8 Max (0803) | Qwen (Alibaba) | 56.2% | Defaultthinking (default reasoning_effort=xhigh; eval setting not stated) | Maker's own figureVendor-reportedQwen (Alibaba) ↗ | — | 3 Aug 2026 |
| 14 | Kimi K3 | Moonshot AI (Kimi) | 56.0% | Maxreasoning_effort=max | Maker's own figureVendor-reportedMoonshot AI ↗ | — | 16 Jul 2026 |
| 15 | GLM-5.3-Flash | Z.ai (Zhipu) | 55.3% | Maxreasoning_effort=max (default) | Maker's own figureVendor-reportedZ.ai ↗ | — | 26 Aug 2026 |
| 16 | Claude Sonnet 5 | Anthropic | 54.9% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 28 Sep 2026 |
| 17 | Claude Opus 4.7 | Anthropic | 54.7% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 28 May 2026 |
| 18 | GLM-5.2 | Z.ai (Zhipu) | 54.7% | Defaultthinking (effort not stated) | Maker's own figureVendor-reportedZ.ai ↗ | — | 16 Jun 2026 |
| 19 | Kimi K2.6 | Moonshot AI (Kimi) | 54.0% | Defaultthinking | Maker's own figureVendor-reportedMoonshot AI ↗ | — | 20 Apr 2026 |
| 20 | Qwen3.7 Max | Qwen (Alibaba) | 53.5% | Defaultthinking (vendor recommends 'Reasoning effort is set to xhigh' system prompt for reasoning tasks) | Maker's own figureVendor-reportedQwen (Alibaba) ↗ | — | 16 May 2026 |
| 21 | Gemini 3.1 Pro (Preview) | Google (Gemini / DeepMind) | 53.4% | MaxDeep Think | Maker's own figureVendor-reportedGoogle ↗ | — | 12 Feb 2026 |
| 22 | GLM-5.1 | Z.ai (Zhipu) | 52.3% | Defaultthinking | Maker's own figureVendor-reportedZ.ai ↗ | — | 7 Apr 2026 |
| 23 | GPT-5.5 | OpenAI | 52.2% | Extra highreasoning effort=xhigh | Maker's own figureVendor-reportedOpenAI ↗ | — | 23 Apr 2026 |
| 24 | GPT-5.4 | OpenAI | 52.1% | Extra highreasoning effort=xhigh | Maker's own figureVendor-reportedOpenAI ↗ | — | 23 Apr 2026 |
| 25 | DeepSeek V4 Flash (0731) | DeepSeek | 51.5% | Maxreasoning_effort=max | Maker's own figureVendor-reportedDeepSeek ↗ | — | 13 Aug 2026 |
| 26 | Qwen3.6 Plus | Qwen (Alibaba) | 50.6% | Defaultthinking | Maker's own figureVendor-reportedQwen (Alibaba) ↗ | — | 2 Apr 2026 |
| 27 | Muse Spark | Meta (Meta Superintelligence Labs, Muse) | 50.4% | Defaulteffort not stated | Maker's own figureVendor-reportedMeta ↗ | — | 9 Jul 2026 |
| 28 | DeepSeek V4 Pro (Preview, 0423) | DeepSeek | 48.2% | MaxThink Max | Maker's own figureVendor-reportedDeepSeek ↗ | — | 24 Apr 2026 |
| 29 | Claude Sonnet 4.6 | Anthropic | 46.8% | Maxadaptive thinking, effort=max | Maker's own figureVendor-reportedAnthropic ↗ | — | 30 Jun 2026 |
| 30 | DeepSeek V4 Flash (Preview, 0423) | DeepSeek | 45.1% | MaxThink Max | Maker's own figureVendor-reportedDeepSeek ↗ | — | 24 Apr 2026 |
| 31 | Gemma 4 31B IT | Google (Gemini / DeepMind) | 26.5% | DefaultThinking | Maker's own figureVendor-reportedGoogle ↗ | — | 2 Apr 2026 |
| 32 | Gemma 4 26B A4B IT | Google (Gemini / DeepMind) | 17.2% | DefaultThinking | Maker's own figureVendor-reportedGoogle ↗ | — | 2 Apr 2026 |