Kimi K2.8 Preview
Moonshot AI
- Released
- September 11, 2026
- Data date
- October 7, 2026
- Input
- Text, images, and video
- Output
- Text
Kimi K2.8 Preview brings more efficient reasoning and improvements in coding and agents over Kimi K2.7 Code, according to Moonshot. The preview rolled out in Kimi Code on September 11, 2026, retaining the service's existing kimi-for-coding ID.
With this checkpoint, Kimi Code supports image and video input and a context of up to 1,048,576 tokens. Reasoning effort can be set to Low, High, or Max.
Specifications and access
| Specification | Value and source |
|---|---|
| Provider model ID | kimi-for-coding (unchanged service ID)Source |
| Access | Kimi CodeSource |
| Input modalities | Text, images, and videoSource |
| Output modalities | TextSource |
| Reasoning effort | Low, High, or Max; Max is the default.Source |
| Model ID | Kimi Code rolls out K2.8 Preview under the existing kimi-for-coding ID.Source |
Sources and data date
| Type | Evidence and data date |
|---|---|
| Input | Documented modalities. Retrieved October 7, 2026. |
| Output | Documented modalities. Retrieved October 7, 2026. |
| Research status | Research date October 7, 2026. 2 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not found in the inspected sources | exact-checkpoint-benchmarks Original wording Kimi Code's official documentation names K2.8 Preview, its model ID, reasoning levels and context, but contains no numerical evaluation results. No independent numerical source for this exact checkpoint was verified in this pass. |
| Additional source | Kimi Code model documentation (retrieved October 7, 2026) · Kimi Code model documentation · Editorial description reviewed October 7, 2026 |