AnthropicClaude Sonnet 4.6
Anthropic's prior-generation balanced model (Sonnet 4.6, superseded by Sonnet 5), kept as a lower-cost option. Strong multi-turn checkout flow with wallet payment and identity linking support; higher token usage than Opus due to verbose tool responses.
- Avg tokens per run
- 72,608
- Avg duration
- 33.9s
- Turns to checkout
- 8.6
Scores come from runs on UCP stores, not estimates: every figure is a tool call, a checkout attempt or a store’s answer.
Shopping score breakdown
- Checkout rate40%34.6%
- Cart rate20%47.9%
- Search rate10%79.9%
- Turn efficiency15%43%
- Token efficiency15%0%
19.4% of runs failed with errors.
Top stores
Development stores are hidden.
| Store | Checkout | Cart |
|---|---|---|
| store hidden | 100% | 100% |
| www.everlane.com | 100% | 100% |
| houseofparfum.nl | 75.5% | 85.9% |
| forever21.com | 66.7% | 66.7% |
| store hidden | 62.5% | 62.5% |
| store hidden | 54.5% | 54.5% |
| store hidden | 50% | 50% |
| mr-fothergills.co.uk | 50% | 50% |
| reebok.com | 50% | 50% |
| westwing.de | 50% | 50% |
Known issues
- MediumNudge Dependency
Stops after validation errors 9.3% of the time. Needs orchestrator nudge to continue checkout.
- LowHigh Token Usage
Averages 67K tokens per run — highest among Anthropic models. Verbose tool response handling.
- MediumMeta Stringification
Frequently stringifies meta objects instead of passing as JSON. Requires orchestrator auto-injection for idempotency keys.
Token usage
- Avg per run
- 72,608
- Average of all models
- 77,769
- Against the average
- −7%
- Median (p50)
- 39,627
- p90
- 138,666
Prompt and completion
Daily average tokens per run, last 30 days
Oct 29Total 47,448Prompt 46,696Completion 752
Range
- Min
- 0
- p25
- 19,558
- p50
- 39,627
- p75
- 77,969
- p90
- 138,666
- Max
- 1,935,784
Test Claude Sonnet 4.6 on your store
Start a run to see how Claude Sonnet 4.6 handles your store’s checkout, call by call.
Compare with other models
Rates are checkout rates. Full comparison view