ModelRefs / Kimi K2 by Moonshot — Benchmarks, Pricing & Review (2026)

Kimi K2 by Moonshot — Benchmarks, Pricing & Review (2026)

Kimi K2 (Moonshot): Kimi K2 by Moonshot.. 200K context. Pricing: from $0.00060/1K in. Specs, benchmarks and code examples.

What this reference supports

Kimi K2 is Moonshot AI's large open-weight mixture-of-experts model, released with instruction-tuned artifacts and positioned for agentic tool use and coding. Its very large total parameter count makes serving architecture the central implementation decision, while hosted endpoints trade that burden for provider dependence.

Kimi K2 is attributed to Moonshot in ModelRefs' canonical registry. Tracked modalities: Text input and output. Primary use cases considered on ModelRefs: Agentic tool-calling and coding workloads on self-managed or partner-hosted stacks; Open-weight evaluation programs comparing frontier-scale MoE serving options.

This ModelRefs profile is Provisional and pending review — decision-support material, not a final or universal ranking. Confirm current behavior, access, pricing, limits, licensing, and lifecycle in Moonshot's own documentation, and evaluate Kimi K2 on representative workloads before implementation.

Benchmark & Evaluation

ModelRefs does not yet hold qualifying sourced benchmark evidence for Kimi K2, so its benchmark coverage is incomplete. Treat any benchmark discussion as provisional and confirm results on representative workloads before selecting it.

  • No benchmark score is imported into this editorial record. Provider-reported evaluations support scoped notes only; canonical score records are governed separately with their own provenance.
  • The release repository and model card report provider-run coding and agentic evaluations; runtime differences mean results do not transfer automatically.

Implementation considerations

  • Size infrastructure to the total MoE parameter count, not the active-parameter figure; expert routing dominates memory and throughput planning.
  • Review the release license terms on the model card and preserve the documented chat template on the chosen runtime.
  • Model artifacts are distributed via Hugging Face and the Moonshot AI repository for licensed deployment.
  • Moonshot and third-party hosts expose separate endpoints, quotas, and commercial terms; map the exact artifact version before comparing providers.

Risks and limitations

  • Open-weight results depend on the exact runtime, precision, quantization, and prompt template; reference results do not transfer automatically.
  • The release-specific license and acceptable-use policy must be reviewed before commercial deployment.

Source coverage

This reference is Provisional. Model behavior, access, pricing, limits, and lifecycle can change; verify the linked provider documentation and run task-specific evaluations before implementation.

Known coverage gaps:

  • Runtime-specific throughput, memory, and total-cost evidence is incomplete.
  • Independent evaluation reproduction is not attached.

Sources

Continue your research

Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Kimi K2 by Moonshot — Benchmarks, Pricing & Review (2026).