| Claude Haiku 4.5 | Qwen3.8 Max (0902) | |
|---|---|---|
| Input, per 1M tokens | $1.00 | $2.00 |
| Output, per 1M tokens | $5.00 | $6.00 |
| Context window | 200K | 1M |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $0.65 | $1.26 |
| Chat reply (2,000 in, 400 out), per 1,000 | $4.00 | $6.40 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $10.50 | $19.00 |
| Confidence signal | no | yes |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | 301 | not measured |
Which one should carry your traffic?
If Claude Haiku 4.5 answers your requests as well as Qwen3.8 Max (0902), moving that traffic saves 38% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Claude Haiku 4.5 does not expose a confidence signal through OpenRouter, so Gimbix would test it as a full replacement rather than a first answer.
Claude Haiku 4.5 profile · Qwen3.8 Max (0902) profile · Measure it on your traffic