GPQA Diamond

Reasoning

Graduate-level science Q&A — 198 expert-validated questions in biology, physics, chemistry

Source ↗ Higher is better

Rank	Model	Provider	Score
1	GPT-6 Astra (xhigh)	OpenAI	96.3%
2	GPT-6 Astra (max)	OpenAI	96.1%
3	Gemini 3.8 Flash (high)	Google	95.3%
4	GPT-6 Astra (high)	OpenAI	94.9%
5	Grok 4.6 (high)	SpaceXAI	94.9%
6	Gemini 3.7 Flash (high)	Google	94.5%
7	Muse Spark 1.3 (xhigh)	Meta	94.1%
8	GPT-5.6 Sol (max)	OpenAI	94.1%
9	Google: Gemini 3.1 Pro Preview	Google	94.1%
10	GPT-6 Astra (medium)	OpenAI	93.9%
11	Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)	Anthropic	93.7%
12	Claude Opus 5 (Adaptive Reasoning, High Effort)	Anthropic	93.7%
13	Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)	Anthropic	93.7%
14	GPT-5.5 (xhigh)	OpenAI	93.5%
15	Grok 4.6 (medium)	SpaceXAI	93.5%
16	Muse Spark 1.3 (max)	Meta	93.5%
17	Qwen: Qwen3.8 2.4T A95B	Alibaba	93.5%
18	Grok 4.6 (xhigh)	SpaceXAI	93.5%
19	Kimi K3 (max)	Moonshot AI	93.5%
20	Gemini 3.8 Flash (medium)	Google	93.5%
21	Claude Fable 5.1 (Adaptive Reasoning, Xhigh Effort, Default Fallback)	Anthropic	93.4%
22	Claude Opus 5 (Adaptive Reasoning, Max Effort)	Anthropic	93.2%
23	GPT-5.5 (high)	OpenAI	93.2%
24	GPT-5.6 Sol (xhigh)	OpenAI	93.1%
25	Grok 4.5 (high)	SpaceXAI	93.1%
26	GPT-6 Astra (low)	OpenAI	93.1%
27	MiniMax: MiniMax M3	MiniMax	92.9%
28	Qwen: Qwen3.8 Max (0902)	Alibaba	92.8%
29	DeepSeek V4 Pro 0813 (Reasoning, Max Effort)	DeepSeek	92.8%
30	GPT-5.6 Sol (high)	OpenAI	92.8%
31	Gemini 3.6 Flash (high)	Google	92.8%
32	Qwen3.8 Max	Alibaba	92.7%
33	Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)	Anthropic	92.6%
34	GPT-5.5 (medium)	OpenAI	92.6%
35	GPT-5.6 Sol (medium)	OpenAI	92.6%
36	GPT-5.6 Terra (max)	OpenAI	92.5%
37	Qwen: Qwen3.7 Max	Alibaba	92.3%
38	Qwen3.8-Flash-Next	Alibaba	92.3%
39	Gemini 3.5 Flash (high)	Google	92.2%
40	Gemini 3.5 Flash (medium)	Google	92.1%
41	Gemini 3.7 Flash (medium)	Google	92.1%
42	GPT-5.4 (xhigh)	OpenAI	92.0%
43	Gemini 3.8 Flash (low)	Google	92.0%
44	Claude Opus 4.8 (Adaptive Reasoning, Max Effort)	Anthropic	92.0%
45	Claude Opus 5 (Adaptive Reasoning, Medium Effort)	Anthropic	91.9%
46	GLM-5.3 (max)	Z.ai	91.7%
47	GPT-5.3 Codex (xhigh)	OpenAI	91.5%
48	Claude Opus 4.7 (Adaptive Reasoning, Max Effort)	Anthropic	91.4%
49	DeepSeek V4 Flash Vision (Reasoning, Max Effort)	DeepSeek	91.3%
50	Z.ai: GLM 5.3 Flash	Z.ai	91.2%

Access this data

JSON GET /api/v1

[
  {
    "rank": 1,
    "model": "GPT-6 Astra (xhigh)",
    "provider": "OpenAI",
    "score": "96.3"
  },
  {
    "rank": 2,
    "model": "GPT-6 Astra (max)",
    "provider": "OpenAI",
    "score": "96.1"
  },
  {
    "rank": 3,
    "model": "Gemini 3.8 Flash (high)",
    "provider": "Google",
    "score": "95.3"
  }
]

Curl

$ curl -s https://rightrank.ai/api/v1/rankings/gpqa-diamond.json | jq '.[:3]'

CLI brew install rightrank

$ rightrank top gpqa-diamond --limit 5
$ rightrank compare claude-opus-4.6 gpt-5.4-pro --category gpqa-diamond
$ rightrank value gpqa-diamond  # best performance per dollar

Claude Code Skill Skill

# In Claude Code, use the RightRank skill:
/rightrank best model for gpqa-diamond
/rightrank compare top 5 by value
/rightrank cheapest model scoring above 80 on gpqa-diamond