| gpt-oss-20b | Qwen3.7 Flash | |
|---|---|---|
| Input, per 1M tokens | $0.018 | $0.03 |
| Output, per 1M tokens | $0.09 | $0.13 |
| Context window | 131K | 1M |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $0.0117 | $0.0193 |
| Chat reply (2,000 in, 400 out), per 1,000 | $0.072 | $0.112 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $0.189 | $0.305 |
| Confidence signal | yes | yes |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | 313 | 286 |
Which one should carry your traffic?
If gpt-oss-20b answers your requests as well as Qwen3.7 Flash, moving that traffic saves 36% on chat-shaped requests. Averages cannot tell you whether it does for your requests. gpt-oss-20b exposes a confidence signal, so Gimbix can let it answer first and hand unsure cases to Qwen3.7 Flash.
gpt-oss-20b profile · Qwen3.7 Flash profile · Measure it on your traffic