UnifyBench

RankingGPQA Diamond

GPQA Diamond

Data updated 22 Oct 2026

Bucket
Hard reasoning
Unit
percent
Direction
Higher is better
Version
diamond
Display harness
GPQA Diamond reported
Board
https://github.com/idavidrein/gpqa

Compare published benchmark results with category weights →

Models

Results are ordered by score in the display harness, followed by models with no result. Other harnesses are listed separately below. Missing results remain unknown.

1–25 of 471 entries

ModelScoreEvidenceSource-recorded date
GPT-6 AstraOpenAI96.061%independent repro2026-10-20
Gemini 3.8 FlashGoogle95.253%independent repro2026-10-20
Grok 4.6xAI94.949%independent repro2026-10-20
Gemini 3.7 FlashGoogle94.545%independent repro2026-10-20
Gemini 3.1 ProGoogle94.141%independent repro2026-09-25
GPT-5.6 SolOpenAI94.141%independent repro2026-10-20
Claude Fable 5.1Anthropic93.737%independent repro2026-10-20
Kimi K3Moonshot93.535%independent repro2026-10-20
Muse Spark 1.3Meta93.535%independent repro2026-10-13
Qwen3.8-2.4T-A95BQwen93.535%independent repro2026-09-17
Claude Opus 5Anthropic93.232%independent repro2026-09-29
MiniMax-M3MiniMax92.929%independent repro2026-10-16
DeepSeek V4-Pro 0813DeepSeek92.828%independent repro2026-09-21
Qwen 3.8-MaxQwen92.828%independent repro2026-10-02
Qwen3.8 Max 0902Qwen92.828%independent repro2026-10-20
Claude Fable 5Anthropic92.626%independent repro2026-09-18
GPT-5.6 TerraOpenAI92.525%independent repro2026-09-21
GLM-5.3Z.ai91.717%independent repro2026-10-20
GLM-5.3-FlashZ.ai91.212%independent repro2026-10-20
GPT-5.6 LunaOpenAI91.111%independent repro2026-09-29
Qwen3.8-27BQwen90.505%independent repro2026-10-20
NVIDIA Nemotron 3 UltraNVIDIA86.667%independent repro2026-10-13
Gemini 3.5 Flash-LiteGoogle83.838%independent repro2026-09-30
gpt-oss-120bOpenAI78.182%independent repro2026-09-21
Mistral Medium 3.5Mistral74.848%independent repro2026-10-06