Benchmarks · reasoning
Roboflow Visual Reasoning
Roboflow Vision Evals: image questions requiring arithmetic, comparisons, spatial relations and multi-step inference (the answer is never written in the image). LLM-judge accuracy, mean of runs.
59 models · 110 results · data as of 2026-11-06 · playground.roboflow.com ↗
Leaderboard
Gemini 4 Argon high
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
1 published measuresOutput tokens per task110 results · lower is better
44 more models under All models in settings.
All results
110 rows| # | Model | Effort | Score | Subscription $ / task | API $ / task | Output tokens |
|---|---|---|---|---|---|---|
| 1 | Gemini 4 Argon | high | 91.4% | $0.009 | $0.013 | 2.2k |
| 2 | GPT-6 Astra | high | 91.2% | $0.001 | $0.021 | 1.4k |
| 3 | Gemini 4 Argon | low | 89.4% | $0.002 | $0.002 | 1.2k |
| 4 | GPT-6.1 Sol | high | 88.7% | $0.000 | $0.004 | 1.4k |
| 5 | GPT-6 Astra | low | 87.2% | $0.001 | $0.016 | 1.3k |
| 6 | GPT-6.1 Terra | high | 87.0% | $0.000 | $0.002 | 1.5k |
| 7 | Opus 5.5 | high | 85.9% | $0.000 | $0.011 | 1.7k |
| 8 | GPT-6.1 Terra | low | 85.4% | $0.000 | $0.002 | 1.4k |
| 9 | Muse Wonder (Max) | low | 85.2% | $0.016 | $0.024 | 2.3k |
| 10 | Gemini 3.8 Flash | high | 84.5% | $0.001 | $0.023 | 6.5k |
| 11 | Sonnet 5.5 | high | 83.9% | $0.000 | $0.006 | 1.7k |
| 12 | GPT-6.1 Sol | low | 83.7% | $0.000 | $0.003 | 1.4k |
| 13 | Muse Wonder (Max) | high | 83.4% | $0.031 | $0.047 | 3.4k |
| 14 | Opus 5.5 | low | 83.0% | $0.000 | $0.009 | 1.6k |
| 15 | Gemini 3.5 Flash | low | 82.1% | $0.008 | $0.008 | 1.9k |
| 16 | Gemini 3.7 Flash | high | 81.9% | $0.001 | $0.005 | 2.3k |
| 17 | Gemini 3.8 Flash | low | 81.2% | $0.000 | $0.004 | 1.8k |
| 18 | Gemini 3.5 Flash | high | 81.0% | $0.018 | $0.018 | 2.9k |
| 19 | Gemini 3.6 Flash | high | 81.0% | $0.000 | $0.009 | 3.3k |
| 20 | Gemini 3.7 Flash | low | 80.8% | $0.000 | $0.002 | 1.5k |
| 21 | Qwen 3.8 Max | high | 80.3% | $0.001 | $0.011 | 2.5k |
| 22 | Gemini 3.6 Flash | low | 77.7% | $0.000 | $0.003 | 1.7k |
| 23 | GPT-6 Sol | high | 77.7% | $0.000 | $0.006 | 1.6k |
| 24 | Muse Spark 1.1 | high | 76.6% | $0.013 | $0.013 | 4.1k |
| 25 | Sonnet 5.5 | low | 76.4% | $0.000 | $0.005 | 1.6k |
| 26 | Qwen 3.8 Max | low | 75.9% | $0.001 | $0.005 | 1.5k |
| 27 | Muse Spark 1.2 | high | 75.7% | $0.006 | $0.012 | 3.9k |
| 28 | Muse Spark 1.2 | low | 75.1% | $0.004 | $0.007 | 2.7k |
| 29 | Haiku 5.5 | high | 74.8% | $0.000 | $0.001 | 2.3k |
| 30 | Gemini 3.1 Pro | high | 74.8% | $0.021 | $0.021 | 2.7k |