Humanity's Last Exam Diamond (HLE-Diamond)
We show this table for reference; we do not rank on it.
Accuracy on the 1,000-question HLE-Diamond set without tools, using the benchmark owner's high-reasoning evaluation.
Accuracy without tools at high reasoning effort on HLE-Diamond — October 5, 2026 source observation
We mirror the published accuracy without tools at high reasoning effort view for HLE-Diamond. GPT-6 Astra leads the public snapshot at 59.9%, followed by Gemini 4 Argon (56.9%) and Claude Opus 5.5 (54.6%). We do not use these results to rank models overall.
GPT-6 Astra
OpenAI
No tools · high reasoning
Gemini 4 Argon
No tools · high reasoning
Claude Opus 5.5
Anthropic
No tools · high reasoning
18 modelsKnowledgeCurrentDisplay onlyUpdated October 5, 2026 source observation
Accuracy without tools at high reasoning effort table (18 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | GPT-6 AstraOpenAINo tools · high reasoning | 59.9% | Not reported | Closed |
| 2 | Gemini 4 ArgonGoogleNo tools · high reasoning | 56.9% | Not reported | Closed |
| 3 | Claude Opus 5.5AnthropicNo tools · high reasoning | 54.6% | Not reported | Closed |
| 4 | GPT-6.1 SolOpenAINo tools · high reasoning | 53.2% | Not reported | Closed |
| 5 | Claude Fable 5.1AnthropicNo tools · high reasoning | 51.6% | Not reported | Closed |
| 6 | Claude Opus 5AnthropicNo tools · high reasoning | 38.1% | Not reported | Closed |
| 7 | Muse WonderMetaNo tools · high reasoning | 34.7% | Not reported | Closed |
| 8 | Gemini 3.8 FlashGoogleNo tools · high reasoning | 33.3% | Not reported | Closed |
| 9 | Claude Sonnet 5.5AnthropicNo tools · high reasoning | 33.1% | Not reported | Closed |
| 10 | GPT-6 SolOpenAINo tools · high reasoning | 32.8% | Not reported | Closed |
| 11 | GPT-5.6 SolOpenAINo tools · high reasoning | 30.4% | Not reported | Closed |
| 12 | Kimi K3.1 ProMoonshot AINo tools · high reasoning | 29.5% | Not reported | Open |
| 13 | Muse Spark OpenMetaNo tools · high reasoning | 24.9% | Not reported | Open |
| 14 | Muse Spark 1.3MetaNo tools · high reasoning | 24.6% | Not reported | Closed |
| 15 | Grok 4.7xAINo tools · high reasoning | 22.8% | Not reported | Closed |
| 16 | GLM 5.3Z.AINo tools · high reasoningText-only subset (728 questions) | 22.3% | Not reported | Unknown |
| 17 | Kimi K3Moonshot AINo tools · high reasoning | 21.7% | Not reported | Open |
| 18 | DeepSeek V4 Pro 0813DeepSeekNo tools · high reasoning | 13.4% | Not reported | Open |
Results reported by model developersUnranked
Labs ran these themselves. They take no rank here.
| Model | Reported value | What was measured | Against the official board | Who ran it |
|---|---|---|---|---|
| Kimi K3.1 ProMoonshot AI | 39.4%Accuracy | 1,000-question set without tools, max effort | Same measure as the board score, run by the lab | Reported by Moonshot AI |
A lab’s own result is shown for reference. It never takes a rank on this board.
How we label sourcesHow to read this leaderboard
HLE-Diamond is a 1,000-question refined subset of Humanity's Last Exam. The published no-tools table evaluates each listed model at high reasoning effort.
Variant check: This is a different question set from original HLE, HLE-Rolling, and HLE-Verified. Highest-effort no-tools and web-and-code runs have separate tables.
Compare rows within this no-tools table. The equal-sized reasoning and knowledge partitions make the overall score an average across two different kinds of question.
Operator receipt: 18 sourced rows are currently displayable on this page; the leading published row is GPT-6 Astra at 59.9%.
Provenance: The scores mirror the Center for AI Safety and Scale AI release table. We have not rerun the gated question set.
Source freshness: HLE-Diamond was released on September 22, 2026. The current release table was checked on October 5, 2026; it does not date each score revision.
Honest limit: The dataset requires access approval, and the release does not publish uncertainty intervals. GLM 5.3 reports a text-only subset. DeepSeek V4 Pro has inconsistent effort labels across its overall score and partition table, so neither row appears in this full-set mirror.
How to read the HLE-Diamond no-tools table
The HLE-Diamond release table checked on October 5, 2026 reports 17 full-set results without tools at high reasoning effort. The 1,000-question subset contains 500 reasoning and 500 knowledge questions. GLM 5.3 ran only the 728 text-only questions (347 reasoning, 381 knowledge), so that score is not a full-set result.
We keep these benchmark-owner results separate from full-set HLE, the highest-effort no-tools view, and web-and-code runs. This table is display only and does not move model rankings.
Snapshot
The published HLE-Diamond snapshot places GPT-6 Astra first at 59.9%. The third row is 5.3 points behind. The broader top-10 range is 27.1 points, so the table still separates the published systems.
18 models have been evaluated on HLE-Diamond. The benchmark falls in the Knowledge category. HLE-Diamond is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About HLE-Diamond
Year
2026
Tasks
1,000 expert questions: 500 reasoning and 500 knowledge
Format
Accuracy without tools, reasoning high
Difficulty
Frontier expert knowledge and reasoning
The release table checked on October 5, 2026 has 12 full-set results without tools at high reasoning effort. HLE-Diamond contains 500 reasoning and 500 knowledge questions drawn from a refined HLE subset. Highest-effort and web-and-code results occupy separate tables.
Freshness and provenance
Version
HLE-Diamond, reasoning high without tools
Refresh cadence
Static
Staleness state
Current
Question availability
Gated 1,000-question dataset; public aggregate results
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does HLE-Diamond measure?
Accuracy on the 1,000-question HLE-Diamond set without tools, using the benchmark owner's high-reasoning evaluation.
Which model leads the published HLE-Diamond snapshot?
GPT-6 Astra currently leads the published HLE-Diamond snapshot with 59.9% accuracy without tools at high reasoning effort. BenchLM shows this benchmark for display only and does not use it in overall rankings.
How many models are evaluated on HLE-Diamond?
The October 5, 2026 source observation snapshot contains 18 AI models.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 7,000+ subscribers.
One email each week. Unsubscribe anytime.