Per-model temperature

Set a temperature override for a model from its provider settings. Kora applies it whenever that model is selected. The available range is 0.0–2.0, in steps of 0.1.

When to adjust

Use case	Recommended
Refactoring, code review	0.2 — deterministic, repeatable
Bug fixing, data extraction	0.3 — precise, focused
Writing, documentation	0.5 — balanced
Brainstorming, ideation	0.7 — creative, varied
Creative writing, poetry	1.0 — maximum creativity

Set an override

The controls live under Settings → [provider] → Model Temperature Overrides. Each model row has a slider, a numeric input, and a reset button.

1. Open Settings and select the provider.
2. Expand Model Temperature Overrides.
3. Adjust the slider or type a value (0–2, one decimal) for the target model.
4. The override applies to the next message. Use the reset (×) button to clear it and fall back to the Smart Token Engine settings.

Precedence

Per-model overrides win over the global temperature override and over the Smart Token Engine mode preset. The full order, from highest to lowest:

1. Per-model temperature override
2. Global temperature override (Smart Token Engine)
3. Mode preset (Eco / Balanced / Quality)

Note

A cyan Temperature badge in the chat indicates when a per-model override is active for the current model.

Troubleshooting

• Override has no effect → the model ID in the override must match the one currently selected; check the provider’s model list.
• Value resets to a different number → inputs are clamped to 0–2 and rounded to one decimal.
• Provider ignores temperature 0 → some providers clamp to a minimum of 0.01 or 0.1 server-side; this is not a Kora bug.
• Badge never shows → no override is set for the active model; set one under Model Temperature Overrides.