Benchmarks · knowledge
GPQA Diamond (Vals)
Graduate-level Google-Proof Q&A benchmark evaluating models on questions that require deep reasoning. Scores are the values published by VALS AI. Source benchmark version 1.
55 models · 55 results · updated 2026-09-01 · vals.ai ↗
Leaderboard
GPT-5.6 Sol max
Heavy line: best score at each price. Thin lines join one model's effort levels. Cost uses each model's median subscription plan, or API price with no plan data. 36 more models under All models in settings.
More from this benchmark
3 published measuresTime per task138 results · lower is better
38 more models under All models in settings.
Score vs any published measure2 measures: sub-scores, splits, counts
Breakdown
darker is better within each column · click a column to sort| Model | ||
|---|---|---|
| GPT-5.6 Sol max | 94.9 | 95.5 |
| Grok 4.6 high | 93.9 | 95.5 |
| Gemini 3.8 Flash high | 94.9 | 93.9 |
| Gemini 3.7 Flash high | 95.5 | 92.4 |
| Qwen 3.8 Max | 92.4 | 94.9 |
| Fable 5.1 max | 92.9 | 93.9 |
| Opus 5 max | 92.4 | 94.4 |
| Gemini 3.6 Flash high | 92.4 | 94.4 |
| Fable 5 max | 92.9 | 93.4 |
| GPT-5.5 xhigh | 93.4 | 92.9 |
| Grok 4.5 high | 92.9 | 92.9 |
| Kimi K3 max | 93.4 | 92.4 |
| Gemini 3.5 Flash high | 91.4 | 93.9 |
| MiniMax M3 max | 92.4 | 92.9 |
| Opus 4.8 max | 91.9 | 92.9 |
| DeepSeek V4 Pro 0813 max | 91.9 | 92.9 |
| GPT-5.6 Luna max | 92.4 | 90.9 |
| Grok 4.3 xhigh | 92.9 | 89.9 |
| Muse Spark 1.1 xhigh | 90.4 | 91.9 |
| GPT-5.6 Terra xhigh | 89.9 | 91.9 |
| Opus 4.7 max | 90.9 | 89.4 |
| Qwen 3.7 Max | 89.9 | 90.4 |
| DeepSeek V4 Flash 0731 high | 89.9 | 89.9 |
| Muse Spark | 88.4 | 90.9 |
| Kimi K2.6 max | 89.9 | 88.4 |
All results
55 rows| # | Model | Effort | GPQA Diamond (Vals) | Subscription $ / task | API $ / task | Time |
|---|---|---|---|---|---|---|
| 1 | GPT-5.6 Sol | max | 95.2% | $0.003 | $0.11 | 58s |
| 2 | Grok 4.6 | high | 94.7% | $0.005 | $0.045 | 3.5 min |
| 3 | Gemini 3.8 Flash | high | 94.4% | $0.002 | $0.056 | 20s |
| 4 | Gemini 3.7 Flash | high | 93.9% | $0.003 | $0.023 | 8.8s |
| 5 | Qwen 3.8 Max | max | 93.7% | $0.009 | $0.077 | 4.4 min |
| 6 | Fable 5.1 | max | 93.4% | $0.032 | $0.20 | 46s |
| 7 | Opus 5 | max | 93.4% | $0.002 | $0.095 | 40s |
| 8 | Gemini 3.6 Flash | high | 93.4% | $0.001 | $0.032 | 18s |
| 9 | Fable 5 | max | 93.2% | $0.051 | $0.45 | 1.7 min |
| 10 | GPT-5.5 | xhigh | 93.2% | $0.002 | $0.083 | 1.9 min |
| 11 | Grok 4.5 | high | 92.9% | $0.007 | $0.044 | 3.6 min |
| 12 | Kimi K3 | max | 92.9% | $0.008 | $0.088 | 2.0 min |
| 13 | Gemini 3.5 Flash | high | 92.7% | $0.064 | $0.064 | 28s |
| 14 | MiniMax M3 | max | 92.7% | $0.004 | $0.025 | 4.6 min |
| 15 | Opus 4.8 | max | 92.4% | $0.006 | $0.34 | 2.2 min |
| 16 | DeepSeek V4 Pro 0813 | max | 92.4% | $0.066 | $0.066 | 4.7 min |
| 17 | GPT-5.6 Luna | max | 91.7% | $0.000 | $0.011 | 54s |
| 18 | Grok 4.3 | xhigh | 91.4% | $0.025 | $0.025 | 12 min |
| 19 | Muse Spark 1.1 | xhigh | 91.2% | $0.034 | $0.034 | 75s |
| 20 | GPT-5.6 Terra | xhigh | 90.9% | $0.001 | $0.034 | 37s |
| 21 | Opus 4.7 | max | 90.2% | $0.37 | $0.37 | 3.4 min |
| 22 | Qwen 3.7 Max | max | 90.2% | $0.019 | $0.061 | 2.1 min |
| 23 | DeepSeek V4 Flash 0731 | high | 89.9% | $0.013 | $0.013 | 80s |
| 24 | Muse Spark | 89.6% | $0.003 | $0.003 | 1.6 min | |
| 25 | Kimi K2.6 | max | 89.1% | $0.040 | $0.080 | 9.1 min |
| 26 | Sonnet 5 | max | 88.9% | $0.001 | $0.078 | 63s |
| 27 | Qwen 3.8 27B | xhigh | 88.9% | $0.003 | $0.047 | 1.9 min |
| 28 | GLM 5.3 | max | 88.1% | $0.006 | $0.056 | 3.5 min |
| 29 | Qwen 3.6 Plus | max | 87.4% | $0.011 | $0.023 | 2.0 min |
| 30 | Inkling | max | 87.1% | $0.015 | $0.17 | 5.1 min |