Nearly every time i use gemma-4-31b-it with Opencode, it just stops mid-chat. Im also using Gemini 3.5 Flash and 3.1 Flash Lite and some other Gemini models, and they dont stop until they are done. Google explain?
This is probably the issue
Hi @NotTHT
Apologies for inconvenience . The chat template was updated a week ago, and we have also reduced model laziness. We have minimized edge cases where the model was holding back or cutting answers short, leading to more complete responses. Could you please try again and let us know if the issue persists?
ref :: Google Gemma on X: "👁️ Vision Options: Want to make Gemma see even better? The default vision bucket is 280 for token efficiency. To capture maximum detail (like sharp OCR and 2.51MP resolution), manually bump max_soft_tokens to 1120! Try our new interactive Space to see how it works:" / X
Thanks