llms.help
Search/
MO

Model profile

Kimi K3-256k

Built by Moonshot AI

The 256K context version of Kimi K3, effectively reducing consumption while delivering same results within 256k context. Ideal for everyday Q&A, code completion, routine feature development, and single-file or small-file edits. Video input not supported.[1]

Context window256,000 tokens[1]
Directory statusActive
Official model cardVisit source ↗

Developer snapshot

Key specifications

Verified Aug 18, 2026

Model identifier
k3-256k[1]
Input modalities
text, image[1]
Output modalities
text[1]
Total parameters
2.8T[1]

Technical profile

What developers should know

01

Intended uses and limits

Everyday Q&A, code completion, routine feature development, single-file or small-file edits where 256k context suffices.[1]

Fixed 256K context limit; no video input support; switching from 1M model when context exceeds 256k requires compact; available to all Moderato members and above.[1]

02

Deployment notes

Available to all Moderato members and above. Use model ID 'k3-256k' in API calls. Base URLs: OpenAI compatible https://api.kimi.com/coding/v1, Anthropic compatible https://api.kimi.com/coding/. When switching from k3 (1M) to k3-256k, if session context exceeds 256k, compact first. Direct switch fails if conversation includes video files. Switching from k3-256k to k3 (1M) does not affect cache.[1]

Choose the artifact

Official checkpoints

Use the exact provider model identifier shown below when selecting this model in an API or compatible SDK.[1]

Official model identifier k3-256k

Provider-documented identifier for this model.[1]