← Scoreboard
DeepSeek V4.1 Flash
Each provider gets the same short prompts. We count how often its top answer is the same as DeepSeek's own.
Tested
10
of 50 on OpenRouter
Match
10
Watching
0
0 no match
Typical agreement
93%
same answer as the maker
| Provider | Same answer as the maker · typical | Result | Precision | Price out / 1M |
|---|---|---|---|---|
| DeepSeek | makersame answer twice 98% of the time | Reference | unknown | $0.60 |
| Wafer · us | 94%100 of 106 prompts · typical 92% | Matches | mxfp4 | $0.50 |
| StreamLake · fp8 | 94%100 of 106 prompts · typical 92% | Matches | fp8 | $0.44 |
| DeepInfra · fp8 | 93%99 of 106 prompts · typical 92% | Matches | fp8 | $0.29 |
| DeepInfra · fp8 flex | 92%98 of 106 prompts · typical 92% | Matches | fp8 | $0.24 |
| Morph · fp8 | 92%97 of 106 prompts · typical 93% | Matches | fp8 | $0.40 |
| Relace · fp4 | 92%97 of 106 prompts · typical 93% | Matches | fp4 | $0.50 |
| Particle AI · us | 90%95 of 106 prompts · typical 93% | Matches | unknown | $0.50 |
| Particle AI · us flex | 89%94 of 106 prompts · typical 93% | Matches | unknown | $0.25 |
| Ionstream · flex | 87%92 of 106 prompts · typical 93% | Matches | unknown | $0.25 |
Question by question
106 questions. Each square is one question. Lit means the same answer as DeepSeek. Hover or tap a column.
Results from Nov 11. Short prompts at temperature 0, so settings cannot change the answer. Honest providers still differ a little, so each one is compared with what is typical for the others. Quality checks are not running yet. A provider is marked as no match only after 2 failed checks on different days.