UCP Playground
Start a run
DeepSeekDeepSeek

DeepSeek R1 (Reasoning) Retired

This model is no longer available to run. Its stats below reflect past runs and are kept for reference.

DeepSeek's reasoning model with chain-of-thought. Slow (76s avg) but deliberate — the reasoning traces show it carefully planning each tool call. Moderate error rate (12.5%) partly from JSON formatting issues. Small sample size but shows promise for complex checkout flows where thinking through validation errors matters. Retired: no longer served by the upstream router (14 Nov 2026).

Avg tokens per run
22,398
Avg duration
62.1s
Turns to checkout
6.7
Compare
Checkout3.4Cart7.9Search37.1Turn efficiency66Token efficiency42
Shopping score
23/100Weak

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%3.4%
  • Cart rate20%7.9%
  • Search rate10%37.1%
  • Turn efficiency15%66%
  • Token efficiency15%42%

62.9% of runs failed with errors.

Top stores

Development stores are hidden.

Store Checkout Cart
everlane.com33.3% 33.3%
allbirds.com33.3% 33.3%
houseofparfum.nl20% 40%
store hidden0% 0%
store hidden0% 0%
store hidden0% 0%
store hidden0% 0%
store hidden0% 0%
store hidden0% 0%
store hidden0% 0%

Known issues

  • MediumHigh Latency

    76 seconds average — reasoning overhead adds significant time.

  • MediumJSON Formatting

    12.5% error rate partly from JSON formatting issues in tool arguments.

Token usage

Avg per run
22,398
Average of all models
78,415
Against the average
−71%
Median (p50)
14,787
p90
51,022

Prompt and completion

Prompt 21,195 (95%)Completion 1,203 (5%)

Daily average tokens per run, last 30 days

Nov 9Total 0Prompt 0Completion 0

Range

Min
0
p25
7,254
p50
14,787
p75
31,217
p90
51,022
Max
116,012

Test DeepSeek R1 (Reasoning) on your store

Start a run to see how DeepSeek R1 (Reasoning) handles your store’s checkout, call by call.

Start a run

Compare with other models

Qwen 3.8 Flash 52.6%Llama 3.3 70B 39.3%Gemini 3.5 Flash 37.6%Claude Sonnet 4.6 34.3%Grok 4.3 26.3%Gemini 2.5 Flash 22.3%Claude Opus 4.8 22.3%DeepSeek V3.2 20.4%DeepSeek V4.1 Flash 20%Gemini 3.1 Pro 19.9%Gemini 2.5 Pro 19.5%GPT-4o 14.7%Claude Sonnet 5 13.4%Grok 3 Mini (Reasoning) 11.1%GPT-5.5 11%DeepSeek V4 Flash 10.9%Grok 4.5 10.9%Claude Opus 5 10.8%o4-mini (Reasoning) 8.9%Gemini 3.6 Flash 8.4%DeepSeek V4 Pro 7.5%Gemini 3.8 Flash 5.6%Qwen 3.7 Max 4.8%Kimi K3 3%Llama 4 Maverick 2.2%GPT-6 Luna 0%Gemini 3.7 Flash 0%Muse Spark 1.3 0%QwQ 32B (Reasoning) 0%All models

Rates are checkout rates. Full comparison view