| Claude Sonnet 5 | Ministral 3 8B 2512 | |
|---|---|---|
| Input, per 1M tokens | $2.00 | $0.15 |
| Output, per 1M tokens | $10.00 | $0.15 |
| Context window | 1M | 262K |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $1.30 | $0.0915 |
| Chat reply (2,000 in, 400 out), per 1,000 | $8.00 | $0.36 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $21.00 | $1.27 |
| Confidence signal | no | no |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | 308 | not measured |
Which one should carry your traffic?
If Ministral 3 8B 2512 answers your requests as well as Claude Sonnet 5, moving that traffic saves 96% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Ministral 3 8B 2512 does not expose a confidence signal through OpenRouter, so Gimbix would test it as a full replacement rather than a first answer.
Claude Sonnet 5 profile · Ministral 3 8B 2512 profile · Measure it on your traffic