CtrlAltDebrief

GPQA Diamond benchmark

398 AI models ranked on GPQA Diamond. Each row shows the model’s score next to its cheapest API price per 1M tokens where it has one, so you can weigh quality against cost.

Top model today

Gemini 3.1 Pro Preview (Google) leads at 94%, from $2.00 per 1M input.

Data via arXiv:2311.12022 →

Models ranked

398

on this benchmark

Top score

94%

current leader

Type

Single eval

benchmark

Updated

15 July 2026

last source fetch

Cost vs GPQA Diamond

211 priced models
Most attractive quadrant (cheap + high)
= reasoning model (extended thinking)
0%20%40%60%80%100%$0.01$0.03$0.1$0.3$1$3$10$30COST / 1M INPUT TOKENS (log) →GPQA DIAMOND SCORE ↑Gemini 3.1 Pro PreviewGPT-5.6 SolGPT-5.5Grok 4.5MiniMax M3Claude Fable 5GPT-5.6 TerraGemini 3.5 FlashMiniMax M2.7Gemma 4 31BMimo V2 FlashQwen3 235B A22b Instruct 2507gpt-oss-120bgpt-oss-20bQwen2 5 Coder 7B InstructLlama 3.1 8B Instruct

Each labelled name links to that model’s page. Curated to the cost-efficiency frontier, top scorers and cheapest so labels stay readable; the full field is in the leaderboard table below (and via “select all”). Reasoning models (extended “thinking” before answering) carry a ring around the dot.

Models ranked on GPQA Diamond

highest score first · 398 models
Maker
#
1Gemini 3.1 Pro PreviewGoogle94%$2.00$12.00
2GPT-5.6 SolOpenAI94%$5.00$30.00
3GPT-5.5OpenAI94%$5.00$30.00
4Grok 4.5xAI93%$2.00$6.00
5MiniMax M3MiniMax93%$0.30$1.20
6Claude Fable 5Anthropic93%$10.00$50.00
7GPT-5.6 TerraOpenAI93%$2.50$15.00
8Qwen3.7 MaxAlibaba92%n/an/a
9Gemini 3.5 FlashGoogle92%$1.50$9.00
10Claude Opus 4.8Anthropic92%$5.00$25.00
11GPT-5.4OpenAI92%$2.50$15.00
12GPT-5.3 CodexOpenAI92%n/an/a
13Claude Opus 4.7Anthropic91%$5.00$25.00
14Claude Sonnet 5Anthropic91%$2.00$10.00
15GPT-5.6 LunaOpenAI91%$1.00$6.00
16Grok 4.20 0309 v2Spacexai91%n/an/a
17Kimi K2.6Moonshot91%n/an/a
18Gemini 3 ProGoogle91%$2.00$12.00
19DeepSeek V4 ProDeepSeek91%high$0.43$0.87
20GPT-5.2OpenAI90%$1.75$14.00
21Grok 4.3xAI90%$1.25$2.50
22Qwen3.7 PlusAlibaba90%n/an/a
23GPT-5.2-CodexOpenAI90%$1.75$14.00
24Muse Spark 1.1Other90%$1.25$4.25
25Gemini 3 FlashGoogle90%reasoning: truen/an/a
26Hy3-previewTencent90%n/an/a
27Claude Opus 4.6Anthropic90%reasoning: adaptive$5.00$25.00
28Kimi K2.7 CodeMoonshot90%$0.95$4.00
29GLM 5.2Zhipu89%$1.40$4.40
30Grok Build 0.1 0616Spacexai89%n/an/a
31DeepSeek V4 FlashDeepSeek89%n/an/a
32Qwen3.5 397B A17BAlibaba89%$0.60$3.60
33Nex-N2-ProNex-agi89%n/an/a
34Qwen3.6 Max PreviewAlibaba89%n/an/a
35Grok 4.20 0309Spacexai88%n/an/a
36Muse SparkMeta88%n/an/a
37Qwen3.6 PlusAlibaba88%$0.33$1.95
38Kimi K2.5Moonshot88%$0.50$2.80
39Grok 4xAI88%$3.00$15.00
40Claude Sonnet 4.6Anthropic87%reasoning: adaptive$3.00$15.00
41GPT-5.4 MiniOpenAI87%$0.75$4.50
42MiniMax M2.7MiniMax87%$0.26$0.55
43GPT-5.1OpenAI87%$1.25$10.00
44DeepSeek V3 2 SpecialeDeepSeek87%$0.58$1.68
45MiMo-V2-ProXiaomi87%n/an/a
46GLM 5.1Zhipu87%$1.05$3.50
47Hy3-previewTencent87%n/an/a
48Nemotron 3 UltraNVIDIA87%n/an/a
49Claude Opus 4.5Anthropic87%reasoning: true$5.00$25.00
50MiMo-V2.5-ProXiaomi87%$1.00$3.00
1–50 of 398
Page 1 of 8
Prices are on-demand list rates. Enterprise commitments can be materially lower. Cost per task uses best-effort token usage.

Where a model has no directly-measured score, a lab-claimed or estimated value is shown.

Frequently asked questions

What is the GPQA Diamond benchmark?

A 198-question multiple-choice subset of GPQA covering graduate-level biology, physics and chemistry, written to be 'Google-proof' so answers can't be found by web search. Scored on accuracy: domain PhDs reach about 65%, while skilled non-experts manage only about 34% even with web access.

Which AI model scores highest on GPQA Diamond?

Gemini 3.1 Pro Preview (Google) leads with 94%, from $2.00 per 1M input tokens.

How many models are ranked on GPQA Diamond?

398 models carry a GPQA Diamond score. Scores come from Artificial Analysis; where a model has a live API price it is verified against LiteLLM and OpenRouter, and models with no current price are listed without one.

Other benchmarks

compare the same models on a different eval
Intelligence Indexcomposite indexCoding Indexcomposite indexAgentic Indexcomposite indexHumanity's Last ExamleaderboardMMLU-ProleaderboardMMMU-ProleaderboardLiveCodeBenchleaderboardSciCodeleaderboardTerminalBench HardleaderboardTerminalBench v2.1leaderboardτ²-benchleaderboardτ-bench BankingleaderboardAPEX AgentsleaderboardIT-Bench SREleaderboardGDPvalleaderboardCritPtleaderboardOmniscienceleaderboardIFBenchleaderboardLong-Context Reasoningcomposite indexSWE-bench VerifiedleaderboardSWE-bench ProleaderboardAIME 2026leaderboardARC-AGI-2leaderboardAA-Briefcaseleaderboard
Benchmark scores via Artificial Analysis · pricing verified against LiteLLM + OpenRouter · refreshed daily · updated 15 July 2026

Every weekday

AI moves fast. Here's your debrief.

News, analysis, tools, and more.

For people who build with AI