Skip to content
Model Pulse

Leaderboard

GPQA Diamondofficial site ↗

20 models scored · metric: accuracy. Prices are the cheapest listed across all serving providers. Cost/run and Pts/$ are priced for the heavy reasoning token profile — thinking tokens outnumber the answer several times over.

A benchmark is priced for the workload it exercises, not a fixed input:output blend: an agent loop and a single hard question bill very differently on the same rate card. Profile: 2k in · 1.5k out · 8k reasoning · 0% cache hit · 12k context.

Scores are facts as publicly reported by labs (see report links below and on each model page) and aggregated via models.dev. GPQA Diamond is maintained by its own project — Model Pulse is not affiliated with or endorsed by it.

Best value · top 5 by pts per dollar · heavy reasoning profile
  1. 1.GPT-5.6 Luna25,392
  2. 2.Muse Glimmer 30B10,438
  3. 3.MiMo-V2.5-Pro10,310
  4. 4.GLM-5.28,624
  5. 5.GPT-5.58,461
20 models
#
🥇GPT-6 Astraopenai
96
$10$50$0.4951941.05M
🥈Fugusakana
95.5
————1M
🥉Fugu Ultrasakana
95.5
$5$30$0.2953241.05M
4GPT-5.6 Solopenai
94.6
$2$10$0.0999561.05M
5GPT-5.4 Proopenai
94.4
$27$160$1.57460.01.05M
6Gemini 3.1 Pro Previewgoogle
94.3
$1$6$0.0591,5981.05M
7Claude Opus 4.7anthropic
94.2
$4$20$0.0382,4791.05M
8GPT-5.5openai
93.6
$0.188$1.13$0.0118,4611.05M
9GPT-5.6 Terraopenai
92.9
$1.50$2$0.0224,2231.05M
10GPT-5.4openai
92.8
$0.75$6$0.0591,5861.05M
11Qwen3.8 Max Previewalibaba
92.6
$2$6$0.0611,5181M
12Qwen3.7 Maxalibaba
92.4
$0.825$2.48$0.0253,6711.06M
13GPT-5.6 Lunaopenai
92.3
$0.06$0.370.363¢25,3921.05M
14GLM-5.2zhipuai
91.2
$0.3$1.05$0.0118,6241.05M
15LongCat-2.0meituan
88.9
$0.3$1.20$0.0127,4081.05M
16GPT-5.4 miniopenai
88
$0.375$4$0.0392,271400K
17MiMo-V2.5-Proxiaomi
86.6
$0.4$0.80.840¢10,3101.05M
18MAI-Code-1-Flashmicrosoft
84.6
$0.75$4.50$0.0441,912256K
19Muse Glimmer 30Bmeta
83.5
$0.2$0.80.800¢10,438131K
20GPT-5.4 nanoopenai
82.8
$0.18$1.10$0.0117,6601.05M
1–20 of 20 models

Score reports: openai.com · console.sakana.ai · openai.com · openai.com · alibabacloud.com · qwen.ai — full source URLs are linked on each model page.

Cheapest scorer: GPT-5.6 Luna at $0.06 input /M.