HLE

🏆 Leaderboard

As of August 20, 2026, Claude Opus 5 is #1 for HLE at 64.7%. Ranked by Humanity's Last Exam: 2,500 expert questions designed so search engines fail. Humanity's Last Exam leaderboard with HLE scores and API prices. Compare the hardest public reasoning exam across GPT, Claude, Gemini, and open models.

Updated August 20, 2026282 models33 providers


Claude Opus 5 Anthropic · Proprietary	64.7%	—	92.9%	$30.00
Claude Mythos Preview Anthropic · Proprietary	64.7%	94.6%	—	$60.00
Claude Fable 5 Anthropic · Proprietary	64.5%	—	85.9%	$60.00
GLM-5.3OSS Z AI · Open Source	62.5%	—	—	$5.80
Muse Spark 1.1 Meta · Proprietary	62.1%	—	91.2%	$5.50
DeepSeek-V4-Pro-0813OSS DeepSeek · Open Source	60%	—	92.4%	$5.28
Claude Opus 4.8 Anthropic · Proprietary	57.9%	93.6%	85.3%	$30.00
Claude Sonnet 5 Anthropic · Proprietary	57.4%	—	80.3%	$12.00
GPT-5.5 Pro OpenAI · Proprietary	57.2%	—	—	$540.00
Kimi K3OSS Moonshot AI · Open Source	56%	93.5%	93.5%	$18.00
GLM-5.2OSS Z AI · Open Source	54.7%	91.2%	71.2%	$5.80
Claude Opus 4.7 Anthropic · Proprietary	54.7%	94.2%	86.4%	$30.00
Seed 2.1 Turbo ByteDance · Proprietary	54.6%	—	—	$3.00
Claude Opus 4.6 Anthropic · Proprietary	53.1%	91.3%	88.4%	$30.00
GLM-5.1OSS Z AI · Open Source	52.3%	86.2%	89.9%	$5.80
GPT-5.5 OpenAI · Proprietary	52.2%	93.6%	77.3%	$35.00
Gemini 3.1 Pro Google · Proprietary	51.4%	94.3%	94.4%	$17.50
Kimi K2-Thinking-0905OSS Moonshot AI · Open Source	51%	84.5%	—	$2.47
Kimi K2.5OSS Moonshot AI · Open Source	50.2%	87.6%	—	$3.68
Qwen3.5-27BOSS Qwen · Open Source	48.5%	85.5%	—	$2.70
DeepSeek-V4-Pro-MaxOSS DeepSeek · Open Source	48.2%	90.1%	89.7%	$5.22
Qwen3.5-122B-A10BOSS Qwen · Open Source	47.5%	86.6%	—	$3.60
Qwen3.5-35B-A3BOSS Qwen · Open Source	47.4%	84.2%	—	$2.25
Claude Sonnet 4.6 Anthropic · Proprietary	46.8%	89.9%	78.8%	$18.00
Gemini 3 Pro Google · Proprietary	45.8%	91.9%	91.9%	$14.00
DeepSeek-V4-Flash-MaxOSS DeepSeek · Open Source	45.1%	88.1%	—	$0.42
Qwen3.8 MaxOSS Qwen · Open Source	43.6%	92.6%	92.7%	$8.00
Gemini 3 Flash Google · Proprietary	43.5%	90.4%	89.4%	$3.50
GLM-4.7OSS Z AI · Open Source	42.8%	85.7%	83.3%	$2.80
Qwen3.7 Max Qwen · Proprietary	41.4%	92.4%	90.9%	$5.00
DeepSeek-V3.2OSS DeepSeek · Open Source · via OpenRouter	40.8%	82.4%	—	$0.57
DeepSeek-V4-Flash-0423OSS DeepSeek · Open Source	40.3%	87.4%	—	$0.30
Gemini 3.5 Flash Google · Proprietary	40.2%	—	88.9%	$10.50
Grok-4 xAI · Proprietary	40%	87.5%	87.5%	$18.00
GPT-5.4 OpenAI · Proprietary	39.8%	92.8%	89.4%	$17.50
Nemotron 3 Ultra (550B A55B)OSS NVIDIA · Open Source · via OpenRouter	37.4%	87%	86.1%	$2.70
GPT-5.2 Pro OpenAI · Proprietary	36.6%	93.2%	—	$189.00
Kimi K2.6OSS Moonshot AI · Open Source	36.4%	90.5%	90.8%	$4.93
Qwen3.7-Plus Qwen · Proprietary	34.7%	90.3%	87.9%	$1.60
GPT-5.2 OpenAI · Proprietary	34.5%	92.4%	73.2%	$15.75
MiMo-V2.5-ProOSS Xiaomi · Open Source	34%	66.7%	82.6%	$1.30
Inkling-SmallOSS Thinking Machines · Open Source · via Thinking Machines Lab	31.6%	89.5%	88.5%	$1.50
Qwen3.8-27BOSS Qwen · Open Source	30.8%	89.2%	—	$3.65
Qwen3.6 Plus Qwen · Proprietary	28.8%	90.4%	88.4%	$3.50
Qwen3.5-397B-A17BOSS Qwen · Open Source	28.7%	88.4%	85.9%	$4.20
GPT-5.4 mini OpenAI · Proprietary	28.2%	88%	88%	$5.25
Gemma 4 31BOSS Google · Open Source	26.5%	84.3%	—	$0.54
LongCat-Flash-Thinking-2601OSS Meituan · Open Source	25.2%	80.5%	—	$1.50
Claude Opus 4.5 Anthropic · Proprietary	25.2%	87%	85.5%	$30.00
GPT-5 OpenAI · Proprietary	24.8%	85.7%	86.2%	$11.25

Showing 1–50 of 282 models

Next step

You found the model. Now ship the product.

Auth, billing, and the API layer are already decided. Fork 8 finished AI apps, or have us build the first version with you.

Get the stack — $249 Have us build it with you

The index

All Large Language Models

282 models across 33 providers. Search or jump to a lab — every model page stays linked here.

Cursor

2 models

Composer 2 Composer 2 Fast

xAI

15 models

Grok 4.6 Grok 4.5 Grok 4.3 Grok-4.1 Grok-4.1 Fast Non-Reasoning Grok-4.1 Fast Reasoning Grok-4.1 Thinking Grok 4 Fast Grok Code Fast 1 Grok-4 Fast Non-Reasoning Grok-4 Fast Reasoning Grok-4 Grok-3 Grok-3 Mini Grok-2

DeepSeek

16 models

DeepSeek-V4-Pro-0813 DeepSeek-V4-Flash-0731 DeepSeek-V4-Flash-0423 DeepSeek-V4-Flash-Max DeepSeek-V4-Pro-Max DeepSeek-V3.2 DeepSeek-V3.2 (Non-thinking)DeepSeek-V3.2-Exp DeepSeek-R1-0528 DeepSeek-V3 0324 DeepSeek R1 Distill Llama 70B DeepSeek R1 Distill Qwen 32B DeepSeek-R1 DeepSeek-V3.1 DeepSeek-V3 DeepSeek-V2.5

MiniMax

6 models

MiniMax M3 MiniMax M2.7 MiniMax M2.5 MiniMax M2.1 MiniMax M2 MiniMax M1 80K

AI21 Labs

2 models

Jamba 1.5 Large Jamba 1.5 Mini

Amazon

5 models

Nova 2 Lite Nova 2 Sonic Nova Lite Nova Micro Nova Pro

Baidu

1 models

ERNIE 4.5

ByteDance

2 models

Seed 2.1 Turbo Seed 2.0 Pro

Cohere

3 models

North Mini Code 1.0 Command A+Command R+

IBM

1 models

Granite 3.3 8B Instruct

Inception

1 models

Mercury 2

inclusionAI

1 models

Ling-3.0-flash

LG AI Research

1 models

K-EXAONE-236B-A23B

Meituan

4 models

LongCat-Flash-Lite LongCat-Flash-Thinking-2601 LongCat-Flash-Thinking LongCat-Flash-Chat

Microsoft

5 models

MAI-Code-1.1-Flash Phi 4 Mini Phi-4-multimodal-instruct Phi 4 Phi-3.5-mini-instruct

Moonshot AI

9 models

Kimi K3 Kimi K2.7 Code Kimi K2.7 Code HighSpeed Kimi K2.6 Kimi K2.5 Kimi K2 0905 Kimi K2-Instruct-0905 Kimi K2-Thinking-0905 Kimi K2 Instruct

Nous Research

1 models

Hermes 3 70B

NVIDIA

5 models

Nemotron 3.5 Lightning (30B A3B)Nemotron 3 Ultra (550B A55B)Nemotron 3 Super (120B A12B)Nemotron 3 Nano (30B A3B)Nemotron Nano 9B v2

Poolside

2 models

Laguna S 2.1 Laguna XS 2.1

Sakana AI

1 models

Sakana Namazu

StepFun

1 models

Step-3.5-Flash

Tencent

1 models

Hy3

Thinking Machines

1 models

Inkling-Small

Unisound

1 models

Upstage

1 models

Solar Pro 4

Xiaomi

3 models

MiMo-V2.5-Pro MiMo-V2.5 MiMo-V2-Flash

Which model leads HLE right now?

As of August 20, 2026, Claude Opus 5 by Anthropic is #1 for HLE at 64.7%. Ranked by Humanity's Last Exam: 2,500 expert questions designed so search engines fail. This board also tracks HLE, GPQA, GPQA Diamond, Humanity's Last Exam (With Tools). Next on the same board: Claude Mythos Preview and Claude Fable 5. Related leaders: GPT-5.6 Sol on GPQA at 94.6%; Gemini 3.7 Flash on GPQA Diamond at 94.8%. This hle leaderboard ranks models by HLE. Scores come from public evals. Prices are the live API rates in the table above.

Sources: Humanity's Last Exam (https://agi.safe.ai/); OpenAI API pricing (https://developers.openai.com/api/docs/pricing); Anthropic Claude API pricing (https://platform.claude.com/docs/en/about-claude/pricing); Gemini API pricing (https://ai.google.dev/gemini-api/docs/pricing)

Top 8 for hle. Ranked by Humanity's Last Exam: 2,500 expert questions designed so search engines fail. Input and output are dollars per million tokens.
Rank	Model	HLE	Input /M	Output /M
1	Claude Opus 5	64.7%	$5.00	$25.00
2	Claude Mythos Preview	64.7%	$10.00	$50.00
3	Claude Fable 5	64.5%	$10.00	$50.00
4	GLM-5.3	62.5%	$1.40	$4.40
5	Muse Spark 1.1	62.1%	$1.25	$4.25
6	DeepSeek-V4-Pro-0813	60%	$1.32	$3.96
7	Claude Opus 4.8	57.9%	$5.00	$25.00
8	Claude Sonnet 5	57.4%	$2.00	$10.00

Related LLM leaderboards

LLM Leaderboard Coding Math Agentic Reasoning Writing Science Finance Long Context Vision Open Source

Individual benchmark pages

Rank one eval at a time. All LLM benchmarks.

SWE Bench Humanity's Last Exam (No Tools)Humanity's Last Exam (With Tools)Terminal-Bench 2.0 Terminal-Bench Terminal-Bench 2.1 Terminal-Bench 2 Terminal-Bench 3.0 ARC-AGI-1 Verified Chatbot Arena

HLE FAQ

Who ranks #1 on the HLE leaderboard?

As of August 20, 2026, Claude Opus 5 by Anthropic ranks #1 on HLE at 64.7%. API pricing is $5.00/M input and $25.00/M output.

What are the top models on HLE?

The current HLE ranking as of August 20, 2026 is 1. Claude Opus 5 at 64.7%; 2. Claude Mythos Preview at 64.7%; 3. Claude Fable 5 at 64.5%.

Which hle model is the cheapest?

Nemotron 3.5 Lightning (30B A3B) is the cheapest scored model on this hle leaderboard at $0.05/M input and $0.20/M output ($0.25 blended). Claude Opus 5 still leads HLE at 64.7%.

Should I always pick the #1 HLE model?

Not automatically. Claude Opus 5 leads HLE, but a cheaper scored model can be the better production choice if the quality gap is small. Use the table to weigh HLE against input/output price, context window, and related evals.

How often is the HLE leaderboard updated?

Scores and API prices on this page are refreshed from published evals and provider rates. The snapshot is labeled August 20, 2026. Treat it as a current index, not a one-off blog post.

What is Humanity's Last Exam?

Humanity's Last Exam, or HLE, is a broad expert-level test from the Center for AI Safety. It is meant to stay hard after models saturate MMLU and GPQA.

What is a good HLE score?

Frontier models still sit well below human experts on the full exam. Compare models on this page rather than treating any single percentage as a pass mark.

Does HLE use tools?

Labs sometimes report HLE with tools and HLE without tools. We keep both columns when the score exists, plus GPQA as a second reasoning check.

You found the model. Now ship the product.

All Large Language Models

OpenAI

Anthropic

Google

Cursor

xAI

DeepSeek

MiniMax

Qwen

AI21 Labs

Amazon

Baidu

ByteDance

Cohere

IBM

Inception

inclusionAI

LG AI Research

Meituan

Meta

Microsoft

Mistral

Moonshot AI

Nous Research

NVIDIA

Poolside

Sakana AI

StepFun

Tencent

Thinking Machines

Unisound

Upstage

Xiaomi

Z AI

Which model leads HLE right now?

Related LLM leaderboards

Individual benchmark pages

HLE FAQ

Who ranks #1 on the HLE leaderboard?

What are the top models on HLE?

Which hle model is the cheapest?

Should I always pick the #1 HLE model?

How often is the HLE leaderboard updated?

What is Humanity's Last Exam?

What is a good HLE score?

Does HLE use tools?