| DeepSeek V4 Pro 0423 | Qwen3.8 27B | |
|---|---|---|
| Input, per 1M tokens | $0.955 | $0.42 |
| Output, per 1M tokens | $1.91 | $3.00 |
| Context window | 1.05M | 1M |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $0.592 | $0.282 |
| Chat reply (2,000 in, 400 out), per 1,000 | $2.67 | $2.04 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $8.60 | $4.86 |
| Confidence signal | yes | yes |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | not measured | 264 |
Which one should carry your traffic?
If Qwen3.8 27B answers your requests as well as DeepSeek V4 Pro 0423, moving that traffic saves 24% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Qwen3.8 27B exposes a confidence signal, so Gimbix can let it answer first and hand unsure cases to DeepSeek V4 Pro 0423.
DeepSeek V4 Pro 0423 profile · Qwen3.8 27B profile · Measure it on your traffic