Knowledge

Artificial Analysis public leaderboard

GPQA

Graduate-level, expert-written science questions evaluated by Artificial Analysis.

Published source

Metric

Accuracy

Results shown

612

Publisher

Artificial Analysis

Snapshot

Fetched 2026-09-08T22:57:37.545Z

612 of 612 models

Higher is better

GPT-6 Astra (xhigh)

GPT-6 Astra (max)

Gemini 3.8 Flash (high)

Grok 4.6 (high)

GPT-6 Astra (high)

Gemini 3.7 Flash (high)

Gemini 3.1 Pro Preview

Muse Spark 1.3 (xhigh)

GPT-5.6 Sol (max)

GPT-6 Astra (medium)

Claude Opus 5 (xhigh)

Claude Fable 5.1 (max with fallback)

Claude Opus 5 (high)

GPT-5.5 (xhigh)

Grok 4.6 (medium)

Gemini 3.8 Flash (medium)

Qwen3.8 2.4T A95B

Muse Spark 1.3 (max)

Grok 4.6 (xhigh)

Kimi K3 (max)

Claude Fable 5.1 (xhigh with fallback)

GPT-5.5 (high)

Claude Opus 5 (max)

Grok 4.5 (high)

GPT-6 Astra (low)

GPT-5.6 Sol (xhigh)

MiniMax-M3

Gemini 3.6 Flash

DeepSeek V4 Pro 0813 (max)

GPT-5.6 Sol (high)

Qwen3.8 Max

GPT-5.5 (medium)

GPT-5.6 Sol (medium)

Claude Fable 5 (with fallback)

GPT-5.6 Terra (max)

Qwen3.7 Max

Qwen3.8-Flash-Next

Gemini 3.5 Flash

Gemini 3.5 Flash (medium)

Gemini 3.7 Flash (medium)

Claude Opus 4.8 (max)

GPT-5.4 (xhigh)

Gemini 3.8 Flash (low)

Claude Opus 5 (medium)

GLM-5.3 (max)

GPT-5.3 Codex (xhigh)

Claude Opus 4.7 (max)

DeepSeek V4 Flash Vision (max)

GLM-5.3-Flash

Kimi K2.6

Claude Sonnet 5 (max)

GPT-5.6 Luna (max)

Grok 4.20 0309 v2

GPT-5.5 (low)

GPT-5.6 Terra (xhigh)

Gemini 3 Pro Preview (high)

DeepSeek V4 Flash 0731 (max)

Claude Fable 5.1 (high with fallback)

DeepSeek V4 Pro (high)

Qwen3.8 27B (xhigh)

Agnes 2.5 Pro Beta

Muse Spark 1.2 (xhigh)

GPT-5.2 (xhigh)

Grok 4.3 (high)

Gemini 3.7 Flash (low)

Qwen3.7 Plus

GPT-5.2 Codex (xhigh)

GPT-5.6 Sol (low)

Gemini 3 Flash

Muse Spark 1.1 (xhigh)

Hy3

Claude Opus 4.6 (max)

GPT-5.6 Terra (high)

Kimi K2.7 Code

Grok Build 0.1 0616

GPT-6 Astra (Non-reasoning)

Inkling Small

GPT-5.6 Luna (xhigh)

GLM-5.2 (max)

DeepSeek V4 Flash (max)

Qwen3.5 397B A17B

Nex-N2-Pro

GPT-5.6 Luna (high)

Solar Pro 4

Grok 4.3 (medium)

Claude Opus 5 (low)

Qwen3.6 Max Preview

DeepSeek V4 Pro (max)

Gemini 3 Pro Preview (low)

Claude Fable 5.1 (medium with fallback)

Grok 4.20 0309

Claude Opus 4.7 (Non-reasoning, high)

Muse Spark

Qwen3.6 Plus

Claude Fable 5.1 (low with fallback)

Kimi K2.5

Grok 4.6 (low)

Grok 4

Agnes 2.5 Pro Alpha

GPT-5.4 mini (xhigh)

Claude Sonnet 4.6 (max)

MiniMax-M2.7

GPT-5.1 (high)

K2 Horizon 375B A23B

Inkling

GPT-5.6 Terra (medium)

DeepSeek V3.2 Speciale

GPT-5.4 (low)

MiMo-V2-Pro

Motif 3 (Beta)

GLM-5.1

Nemotron 3 Ultra

DeepSeek V4 Flash (high)

Hy3-preview

MiMo-V2.5-Pro

Claude Opus 4.5

Apodex 1.1

GPT-5.2 (medium)

Qwen3.5 397B A17B (Non-reasoning)

Qwen3 Max Thinking

GPT-5.1 Codex (high)

GPT-5.6 Luna (medium)

GLM-4.7

Qwen3.5 27B

Qwen3.5 122B A10B

Solar Open2 250B

A.X-K2

Gemma 4 31B

Ring-2.6-1T

MiMo-V2-Omni-0327

Ling 3.0 Flash

KAT-Coder-Pro V2

GPT-5 (high)

Grok 4.1 Fast

MiMo-V2.5

Nanbeige4.1-3B

MiniMax-M2.5

GLM-5-Turbo

Grok 4 Fast

MiMo-V2-Flash

GPT-5.5 Instant (May 2026)

Qwen3.5 35B A3B

Qwen3.8 27B (medium)

Qwen3.8 27B (low)

JT-4.1 Flash 236B A21B

o3-pro

Gemini 2.5 Pro

GPT-5.6 Terra (low)

Grok 4.3 (low)

Qwen3.5 27B (Non-reasoning)

Kimi K3 (low)

Qwen3.6 27B

GPT-5 (medium)

Qwen3.6 35B A3B

Claude Opus 4.6 (Non-reasoning, high)

DeepSeek V3.2

GLM-5.1 (Non-reasoning)

Gemini 3.5 Flash-Lite

Kimi K2 Thinking

GPT-5 Codex (high)

Gemini 2.5 Pro (Mar)

GPT-5.6 Luna (low)

MiMo-V2-Flash (Feb 2026)

Muse Glimmer (high)

Claude 4.5 Sonnet

Motif 3

Step 3.5 Flash

MiniMax-M2.1

JT-35B-Flash

K-

K-EXAONE 2.0

Qwen3.6 27B (Non-reasoning)

Gemini 3.5 Flash (minimal)

GPT-5 mini (high)

MiMo-V2-Omni

Qwen3.5 122B A10B (Non-reasoning)

Qwen3.5 Omni Plus

Step 3.5 Flash 2603

GPT-5.4 mini (medium)

GPT-5.5 Instant (June 2026)

Gemini 2.5 Pro (May)

Gemini 3.1 Flash-Lite

GLM-5

Qwen3.5 35B A3B (Non-reasoning)

Qwen3.8 27B

Qwen3.6 35B A3B (Non-reasoning)

GPT-5.4 nano (xhigh)

DeepSeek R1 0528

GPT-5.1 Codex mini (high)

Gemini 3 Flash (Non-reasoning)

Nova 2.0 Lite (high)

ERNIE 4.5 300B A47B

Claude Opus 4.5 (Non-reasoning)

GLM 5V Turbo

Step 3.7 Flash

Claude 4.1 Opus

GPT-5 (low)

Qwen3.5 9B

G9v3-39A5B

GPT-5 mini (medium)

Nemotron 3 Super

Claude Sonnet 5 (Non-reasoning)

Claude Sonnet 4.6 (Non-reasoning, high)

DeepSeek V3.2 Exp

Claude Sonnet 4.6 (Non-reasoning, low)

Claude 4 Opus

EXAONE 4.5 33B

Gemini 2.5 Flash (Sep)

DeepSeek V3.1 Terminus

Gemma 4 26B A4B

Grok 3 mini Reasoning (high)

Gemini 2.5 Flash

GPT-5.6 Sol (Non-reasoning)

Qwen3 235B A22B 2507

Kimi K2.5 (Non-reasoning)

Kimi K2.6 (Non-reasoning)

Qwen3.5 9B (Non-reasoning)

Nova 2.0 Pro Preview (medium)

Grok 4.20 0309 (Non-reasoning)

o4-mini (high)

K-

K-EXAONE

GLM-4.5

gpt-oss-120b (high)

LongCat 2.0

GLM-4.6

DeepSeek V3.1

ERNIE 5.0 Thinking Preview

Claude 4 Sonnet

MiniMax-M2

Grok 4.20 0309 v2 (Non-reasoning)

Qwen3 Max Thinking (Preview)

Ring-1T

o3-mini (high)

Claude 3.7 Sonnet

Qwen3 VL 235B A22B (Reasoning)

Qwen3.5 4B

Mercury 2

Mistral Small 4

GPT-5.5 (Non-reasoning)

Cogito v2.1

Nova 2.0 Lite (medium)

Kimi K2 0905

Kimi K2

Gemini 2.5 Flash (Sep) (Non-reasoning)

o1-preview

Doubao Seed Code

Qwen3 Max (Preview)

Qwen3 Max

KAT-Coder-Pro V1

Gemma 4 31B (Non-reasoning)

MiMo-V2.5-Pro (Non-reasoning)

Command A+

INTELLECT-3

GPT-5.4 nano

Nova 2.0 Omni (medium)

Qwen3 Next 80B A3B (Reasoning)

Nemotron Cascade 2 30B A3B

North Mini Code

Nemotron 3 Nano

Qwen3 235B 2507 (Non-reasoning)

Gemma 4 12B

Ling-2.6-1T

Trinity Large Thinking

DeepSeek V3.2 (Non-reasoning)

DeepSeek V3.1 Terminus (Non-reasoning)

Nova 2.0 Pro Preview (low)

Mistral Medium 3.5

Llama Nemotron Super 49B v1.5

o3-mini

GPT-5.4 (Non-reasoning)

GPT-5.6 Terra (Non-reasoning)

Nemotron 3.5 Lightning

Qwen3.5 Omni Flash

EXAONE 4.0 32B

Magistral Medium 1.2

DeepSeek V3.2 Exp (Non-reasoning)

Qwen3 Next 80B A3B

Sarvam 105B (high)

Qwen3 Coder Next

DeepSeek V3.1 (Non-reasoning)

Ling 3.0 Tiny

HyperNova 60B 2605 (high)

GLM-4.5-Air

Apriel-v1.6-15B-Thinker

Qwen3 VL 32B (Reasoning)

Hy3-preview (Non-reasoning)

Quasar 438B (max)

Llama Nemotron Ultra

Hermes 4 405B

Claude 4.5 Sonnet (Non-reasoning)

Grok Code Fast 1

Seed-OSS-36B-Instruct

Qwen3 Omni 30B A3B (Reasoning)

Ring-flash-2.0

Solar Pro 3

Mi:dm K 2.5 Pro Preview

Qwen3 VL 30B A3B (Reasoning)

Ling-1T

GLM-4.6V

DeepSeek V4 Pro (Non-reasoning)

DeepSeek V4 Flash (Non-reasoning)

Gemma 4 26B A4B (Non-reasoning)

K2 Think V2

Apriel-v1.5-15B-Thinker

Qwen3.5 4B (Non-reasoning)

GPT-5.2 (Non-reasoning)

Qwen3 VL 235B A22B

Gemini 2.5 Flash-Lite (Sep)

DeepSeek R1 (Jan)

Qwen3 30B A3B 2507

MiniCPM5-2B

Gemini 2.0 Flash Thinking exp. (Jan)

Claude 4 Opus (Non-reasoning)

Mi:dm K 2.5 Pro

Qwen3 235B

Hermes 4 70B

Nova 2.0 Omni (low)

Gemini 2.5 Flash (Apr)

Nova 2.0 Lite (low)

MiniMax M1 80k

Qwen3 VL 30B A3B

Motif-2-12.7B

K-

K-EXAONE (Non-reasoning)

Grok 3

Step3 VL 10B

gpt-oss-20b (high)

GPT-5 mini (minimal)

Solar Pro 2

GLM-5.2 (Non-reasoning)

GPT-5 (ChatGPT)

GLM-4.5V

Gemini 2.5 Flash (Non-reasoning)

Claude 4 Sonnet (Non-reasoning)

MiniMax M1 40k

K2-V2 (high)

Mistral Large 3

Magistral Medium 1

JT-MINI

GPT-5 nano (high)

GPT-5 (minimal)

Claude 4.5 Haiku

gpt-oss-120b (low)

Qwen3 VL 32B

Llama 4 Maverick

GPT-5 nano (medium)

DiffusionGemma 26B A4B

Qwen3 32B

Qwen3 4B 2507

GPT-4.1

GLM-5 (Non-reasoning)

GLM-4.7 (Non-reasoning)

GPT-4.1 mini

Magistral Small 1.2

Gemma 4 12B (Non-reasoning)

Falcon-H1R-7B

Qwen3 30B A3B 2507 (Non-reasoning)

Grok 4.3 (Non-reasoning)

Solar Open 100B

Ling-flash-2.0

MiMo-V2-Flash (Non-reasoning)

Claude 3.7 Sonnet (Non-reasoning)

GPT-4o (Mar)

DeepSeek V3 0324

Gemini 2.5 Flash-Lite (Sep) (Non-reasoning)

Claude 4.5 Haiku (Non-reasoning)

GPT-5.6 Luna (Non-reasoning)

Granite 4.2 30B

Llama 3.3 Nemotron Super 49B

GPT-5.1 (Non-reasoning)

Magistral Small 1

Grok 4.1 Fast (Non-reasoning)

LongCat Flash Lite

Nova 2.0 Pro Preview (Non-reasoning)

Gemini 2.0 Flash (exp)

Sarvam 30B (high)

GLM-4.6 (Non-reasoning)

Celeris-1

Granite 4.2 8B

EXAONE 4.0 32B (Non-reasoning)

Gemini 2.5 Flash-Lite

Sonar Reasoning

Gemini 2.0 Flash

Gemini 2.0 Pro Experimental

Qwen3 Omni 30B A3B

Qwen3 Coder 480B

Qwen3 30B

HyperCLOVA X SEED Think (32B)

DeepSeek R1 Distill Qwen 32B

Qwen3 235B (Non-reasoning)

DeepSeek R1 0528 Qwen3 8B

gpt-oss-20b (low)

Olmo 3 32B Think

Grok 4 Fast (Non-reasoning)

GPT-5.4 mini (Non-reasoning)

Qwen3 14B

Nova 2.0 Lite (Non-reasoning)

o1-mini

Tri-21B-Think

Claude 3.5 Sonnet (Oct)

K2-V2 (medium)

Devstral 2

Gemini 2.5 Flash (Apr) (Non-reasoning)

Ling 2.6 Flash

QwQ-32B

Olmo 3.1 32B Think

Gemini 1.5 Pro (Sep)

Qwen3 8B

Mistral Medium 3.1

Llama 4 Scout

Qwen2.5 Max

GLM-4.7-Flash

Qwen3 VL 8B (Reasoning)

Sonar Pro

Mistral Medium 3

Solar Pro 2 (Preview)

Gemma 4 E4B

Phi-4

GLM-4.5V (Non-reasoning)

NVIDIA Nemotron Nano 12B v2 VL

Ministral 3 14B

Mistral Small 4 (Non-reasoning)

NVIDIA Nemotron Nano 9B V2

Nova Premier

GLM-4.6V (Non-reasoning)

Ling-mini-2.0

Solar Pro 2 (Non-reasoning)

Claude 3.5 Sonnet (June)

Granite 4.2 3B

GPT-5.4 nano (Non-reasoning)

LFM2.5-2.6B

NVIDIA Nemotron Nano 9B V2 (Non-reasoning)

DeepSeek V3 (Dec)

QwQ 32B-Preview

Nova 2.0 Omni (Non-reasoning)

Gemma 4 E4B (Non-reasoning)

Solar Pro 2 (Preview) (Non-reasoning)

GPT-4o (Nov)

Gemini 2.0 Flash-Lite (Preview)

K2-V2 (low)

Olmo 3.1 32B Instruct

Tri-21B-think Preview

Hermes 4 405B (Non-reasoning)

Gemini 2.0 Flash-Lite (Feb)

Qwen3 32B (Non-reasoning)

Devstral Small 2

Reka Flash 3

Command A

GPT-4o (May)

Qwen3 4B

GPT-4o (Aug)

Qwen3 4B 2507 (Non-reasoning)

Llama 3.3 Nemotron Super 49B (Non-reasoning)

Olmo 3 7B Think

Qwen3 Coder 30B A3B

Tulu3 405B

Llama 3.1 405B

Exaone 4.0 1.2B

Qwen3 30B (Non-reasoning)

Nemotron 3 Nano 4B

LFM2.5-8B-A1B

GPT-4.1 nano

GPT-4o (ChatGPT)

Grok 2

Mistral Small 3.2

Pixtral Large

Nova Pro

Llama 3.3 70B

Qwen3 VL 4B (Reasoning)

Devstral Medium

Hermes 4 70B (Non-reasoning)

Qwen2.5 72B

Claude 3 Opus

Mistral Large 2 (Nov)

DeepSeek R1 Distill Qwen 14B

Llama Nemotron Super 49B v1.5 (Non-reasoning)

Granite 4.1 30B

Gemini 2.5 Flash-Lite (Non-reasoning)

LFM2 24B A2B

Mistral Large 2 (Jul)

Sonar

Ministral 3 8B

Grok Beta

Qwen3 14B (Non-reasoning)

Nemotron 3 Nano Omni 30B A3B

Qwen2.5 Instruct 32B

Llama 3.1 Nemotron 70B

Gemini 1.5 Flash (Sep)

Mistral Small 3

Qwen3.5 2B

Mistral Small 3.1

Qwen3 8B (Non-reasoning)

GLM-4.7-Flash (Non-reasoning)

NVIDIA Nemotron Nano 12B v2 VL (Non-reasoning)

Qwen3.5 2B (Non-reasoning)

G9v3-3B

Devstral Small (May)

Nova Lite

Granite 4.1 8B

Gemma 4 E2B

Llama 3.2 90B (Vision)

GPT-5 nano (minimal)

Gemma 3 27B

Qwen3 VL 8B

Jamba 1.5 Large

GPT-4o mini

Molmo2-8B

Mistral Saba

Exaone 4.0 1.2B (Non-reasoning)

DeepSeek-V2.5 (Dec)

Qwen2.5 Coder 32B

Sarvam M

Granite 4.0 H Small

Devstral Small

Kimi Linear 48B A3B Instruct

Qwen2.5 Turbo

Llama 3.1 70B

Claude 3.5 Haiku

Llama 3.1 Nemotron Nano 4B v1.1

Gemma 4 E2B (Non-reasoning)

DeepSeek R1 Distill Llama 70B

Hermes 3 - Llama-3.1 70B

Olmo 3 7B

Claude 3 Sonnet

Nemotron 3 Nano (Non-reasoning)

Qwen3 4B (Non-reasoning)

Jamba 1.7 Large

Jamba 1.6 Large

DeepHermes 3 - Mistral 24B

Mistral Small (Sep)

Llama 3 70B

Claude 3 Haiku

Gemini 1.5 Pro (May)

Qwen2 72B

Qwen3 VL 4B

Gemini 1.5 Flash-8B

Nova Micro

Ministral 3 3B

Qwen3 1.7B

Mistral Large (Feb)

Mistral Medium

Gemma 3 12B

GPT-4

LFM2 8B A1B

Claude 2.0

LFM2.5-1.2B-Thinking

Qwen2.5 Coder 7B

Granite 3.3 8B

Granite 4.0 Micro

Jamba Reasoning 3B

Mixtral 8x22B

DBRX

Phi-4 Mini

Claude Instant

OLMo 2 32B

LFM 40B

Llama 2 Chat 70B

LFM2.5-1.2B-Instruct

Gemini 1.5 Flash (May)

Command-R+ (Apr)

Jamba 1.7 Mini

Llama 2 Chat 13B

Claude 2.1

DeepSeek Coder V2 Lite

Phi-3 Mini

Phi-4 Multimodal

Granite 4.1 3B

LFM2 2.6B

MiniCPM-V 4.6 1.3B

Tiny Aya Global

DeepSeek R1 Distill Llama 8B

Mistral Small (Feb)

Jamba 1.5 Mini

Jamba 1.6 Mini

GPT-3.5 Turbo

Llama 3 8B

Gemma 3n E4B

Mixtral 8x7B

Gemma 3 4B

LFM2.5-VL-1.6B

Qwen1.5 Chat 110B

OLMo 2 7B

Command-R (Mar)

Qwen3 1.7B (Non-reasoning)

Granite 4.0 1B

MiniCPM5-1B

Gemma 3n E4B (May)

Gemini 1.0 Pro

Apertus 70B Instruct

DeepHermes 3 - Llama-3.1 8B

MiniCPM5-1B (Non-reasoning)

Granite 4.0 H 1B

Granite 4.0 350M

Llama 3.1 8B

Granite 4.0 H 350M

Apertus 8B Instruct

Llama 3.2 3B

Molmo 7B-D

Qwen3 0.6B

Gemma 3 1B

Qwen3.5 0.8B (Non-reasoning)

Qwen3 0.6B (Non-reasoning)

OpenChat 3.5

Gemma 3n E2B

LFM2 1.2B

Llama 2 Chat 7B

Gemma 3 270M

Llama 3.2 11B (Vision)

Llama 3.2 1B

Mistral 7B

Qwen3.5 0.8B

DeepSeek R1 Distill Qwen 1.5B

List bars visualize relative performance. Plot bars use published 0–100 scores, shrink to a 1.5rem minimum, then scroll horizontally.

Scores are published snapshots and are only comparable within the same benchmark, version, metric, and evaluation configuration.