HLE-Diamond

GPT-6 Astra 65.8% leads since Sep 2026 In use

Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers.

The makers built it after a year of cleaning as a migration path for when the original set hits its noise ceiling. GPT-6 Astra led at 59.9 percent without tools.

Made by Center for AI Safety and Scale AI Released Sep 2026 Expert-written questions All tasks public Last in a launch post Oct 2026 Last model added Oct 2026 Scores refreshed 2 Nov 2026

Level: Research frontier Researchers, on open problems or tasks that take days
Scored: Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers: Short answer A number, word or short string matched to a key.

centerforaisafety/hle 1,737 stars Dataset Leaderboard lastexam.ai

Models scored on its own leaderboard

Launch post lists it, from 1 lab

GPT-6 Astra leads with 65.8%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

Bars Table

18 models. Scroll sideways for the rest.

#	Model	Score	Setting	Date	Source
1	GPT-6 Astra	65.8%	max	Sep 2026	Official leaderboard
2	GPT-6.1 Sol	61.3%	max	Sep 2026	Official leaderboard
3	Claude Opus 5.5	60.9%	max	Sep 2026	Official leaderboard
4	Gemini 4 Argon	56.9%	high	Sep 2026	Official leaderboard
5	Claude Fable 5.1	55.1%	max	Sep 2026	Official leaderboard
6	Claude Sonnet 5.5	43.3%	max	Sep 2026	Official leaderboard
7	GPT-6 Sol	43.2%	max	Sep 2026	Official leaderboard
8	GPT-5.6 Sol	41%	max	Jul 2026	Official leaderboard
9	Muse Wonder	40.6%	max	Oct 2026	Official leaderboard
10	Claude Opus 5	40.5%	max	Jul 2026	Official leaderboard
11	Kimi K3.1 Pro	38.7%	max	Oct 2026	Official leaderboard
12	Muse Spark 1.3	34.6%	max	Sep 2026	Official leaderboard
13	Gemini 3.8 Flash	33.3%	high	Sep 2026	Official leaderboard
14	Kimi K3	28.3%	max	Jul 2026	Official leaderboard
15	Muse Spark Open	28.2%	max	Oct 2026	Official leaderboard
16	GLM-5.3	25.1%	max	Aug 2026	Official leaderboard
17	Grok 4.7	25%	max	Sep 2026	Official leaderboard
18	DeepSeek-V4-Pro	16.6%	max	Apr 2026	Official leaderboard

Labs report scores for 1 model in their cards. We plot every source in the charts below.

More effort adds a median 6 points

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Chart Table

Lowest effort to highest, light to dark

GPT-6 Astra

high max

65.8%

GPT-6.1 Sol

high max

61.3%

Claude Opus 5.5

high max

60.9%

Gemini 4 Argon

high

56.9%

Claude Fable 5.1

high max

55.1%

Claude Sonnet 5.5

high max

43.3%

GPT-6 Sol

high max

43.2%

GPT-5.6 Sol

high max

41%

Muse Wonder

high max

40.6%

Claude Opus 5

high

40.5%

Kimi K3.1 Pro

high max

38.7%

Muse Spark 1.3

high max

34.6%

Model	Effort	Score
GPT-6 Astra	max	65.8%
GPT-6.1 Sol	max	61.3%
Claude Opus 5.5	max	60.9%
GPT-6 Astra	high	59.9%
Gemini 4 Argon	high	56.9%
Gemini 4 Argon	max	56.9%
Claude Fable 5.1	max	55.1%
Claude Opus 5.5	high	54.6%
GPT-6.1 Sol	high	53.2%
Claude Fable 5.1	high	51.6%
Claude Sonnet 5.5	max	43.3%
GPT-6 Sol	max	43.2%
GPT-5.6 Sol	max	41%
Muse Wonder	max	40.6%
Claude Opus 5	max	40.5%
Kimi K3.1 Pro	max	38.7%
Claude Opus 5	high	38.1%
Muse Wonder	high	34.7%
Muse Spark 1.3	max	34.6%
Gemini 3.8 Flash	high	33.3%
Gemini 3.8 Flash	max	33.3%
Claude Sonnet 5.5	high	33.1%
GPT-6 Sol	high	32.8%
GPT-5.6 Sol	high	30.4%
Kimi K3.1 Pro	high	29.5%
Kimi K3	max	28.3%
Muse Spark Open	max	28.2%
GLM-5.3	max	25.1%
Grok 4.7	max	25%
Muse Spark Open	high	24.9%
Muse Spark 1.3	high	24.6%
Grok 4.7	high	22.8%
GLM-5.3	high	22.3%
Kimi K3	high	21.7%
DeepSeek-V4-Pro	max	16.6%
DeepSeek-V4-Pro	high	13.4%

2 models have held number one since its release in Sep 2026

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

Anthropic OpenAI

Own leaderboard 2 leaders

GPT-6 Astra 65.8%

Scores over time

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

On the benchmark’s own leaderboard Reported by the lab in its model card Best so far

One lab put it in a launch post in the last six months

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

Date	Launch post	Score	As the post writes it
15 Oct 2026	Muse Spark Open	26.1	HLE-Diamond (no tools)

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.