| Gemma 4 31B | Hy4 preview | |
|---|---|---|
| Input, per 1M tokens | $0.09 | $0.834 |
| Output, per 1M tokens | $0.34 | $2.50 |
| Context window | 262K | 1.05M |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $0.0574 | $0.525 |
| Chat reply (2,000 in, 400 out), per 1,000 | $0.316 | $2.67 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $0.89 | $7.92 |
| Confidence signal | yes | no |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | 304 | not measured |
Which one should carry your traffic?
If Gemma 4 31B answers your requests as well as Hy4 preview, moving that traffic saves 88% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Gemma 4 31B exposes a confidence signal, so Gimbix can let it answer first and hand unsure cases to Hy4 preview.
Gemma 4 31B profile · Hy4 preview profile · Measure it on your traffic