Atlas
Benchmarks
326 sources · 1,256 models × 2,956 benchmarks · 154,484 scoresVisa Agentic Commerce Benchmark (pass^3)
The share of 15 shopping episodes in which all three of a model's runs passed, from the same runs as the Visa Agentic Commerce Benchmark's pass rate. Visa's whitepaper counts a capability as demonstrated only when it appears consistently across runs.
Models
22
Top score
66.7
Leader
Muse Wonder
Updated
Nov 5, 2026
Resolution
—
Saturation
—
Signal
38thpercentile
480th strongest of 779 fitted benchmarks
Signal-to-noise +4.70 for the Visa Agentic Commerce Benchmark family
Difficulty
Not estimated
Another benchmark represents this family in the fit.
| # | Model | Setting | Score | Source |
|---|---|---|---|---|
| 1 | Muse Wonder | not reported | 66.7 | |
| 2 | Claude Fable 5.1 | not reported | 60.0 | |
| 3 | Gemini 4 Argon | not reported | 60.0 | |
| 4 | Claude Opus 5.5 | not reported | 46.7 | |
| 5 | GPT-6.1 Sol | not reported | 46.7 | |
| 6 | Claude Sonnet 5.5 | not reported | 40.0 | |
| 7 | GLM-5.3 | not reported | 40.0 | |
| 8 | GPT-6 Astra | not reported | 40.0 | |
| 9 | Grok 4.7 | not reported | 40.0 | |
| 10 | Qwen3.8 Max (0902) | not reported | 40.0 | |
| 11 | Beam 1.0 501B A23B | not reported | 33.3 | |
| 12 | GPT-6.1 Terra | not reported | 33.3 | |
| 13 | Kimi K3.1 Pro | not reported | 33.3 | |
| 14 | DeepSeek V4.1 Flash 1017 | not reported | 26.7 | |
| 15 | Gemini 3.8 Flash | not reported | 26.7 | |
| 16 | GPT-6 Luna | not reported | 26.7 | |
| 17 | Muse Spark 1.3 | not reported | 26.7 | |
| 18 | MiniMax M3.1 Flash | not reported | 20.0 | |
| 19 | MiniMax M3 | not reported | 20.0 | |
| 20 | DeepSeek-V4-Pro | not reported | 13.3 | |
| 21 | Gemini 3.1 Pro Preview | not reported | 13.3 | |
| 22 | Inkling | not reported | 0.0 |
