moonshotai
MoonshotAI: Kimi K2 Thinking
moonshotai/kimi-k2-thinking
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Kimi K2 Thinking is Moonshot AI's open-weight reasoning model, cataloged as moonshotai/kimi-k2-thinking. Its model card describes a trillion-parameter Mixture-of-Experts design activating 32B parameters per token, a 256K context window, and native INT4 quantization produced through quantization-aware training. Reasoning is built in rather than optional: the OpenRouter catalog marks it mandatory, and the documentation emphasizes interleaved chain-of-thought with function calls for long agentic runs. Reported results are strong on reasoning, agentic search, and coding tasks, though they come mainly from the publisher.[1][2]
Strengths
- Publisher reports stable, goal-directed tool use across hundreds of sequential calls, with reasoning interleaved between function invocations.[2]
- Large 256K context window plus native INT4 weights; the model card says quantization-aware training preserves accuracy while reducing latency and GPU memory.[1][2]
- Open weights under a Modified MIT license, with deployment guidance for vLLM, SGLang, and KTransformers and API support for tools and structured outputs.[1][2]
Limitations
- Always-on reasoning: the OpenRouter catalog lists it as mandatory with no configurable effort levels, so it cannot be switched off for simple prompts.[1]
- Most performance data is vendor-reported; the card itself cautions that its consumer chat mode may not reproduce benchmark results, and independent measurements in this pack are sparse.[1][2]
- Text-only input and output; knowledge cutoff and minimum serving hardware are not stated in these sources, so self-hosting needs should be validated per workload.[1][2]
Where to use it
- Plausibly suited to agentic research or browsing pipelines that chain many tool calls, based on the publisher's long-horizon claims; worth piloting before production reliance.[2]
- A candidate for coding assistants and software-engineering agents, given documented tool calling and vendor-reported coding evaluations.[1][2]
- Long-document Q&A or multi-step analytical tasks that can use the wide context window and built-in reasoning.[1][2]
Sources
- [1] OpenRouter catalog · MoonshotAI: Kimi K2 ThinkingSources checked: Nov 13, 2026
- [2] Model documentation · MoonshotAI: Kimi K2 ThinkingSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning. Built on the trillion-parameter Mixture-of-Experts (MoE) architecture introduced in...
Specifications
- Model ID
- moonshotai/kimi-k2-thinking
- Canonical ID
- moonshotai/kimi-k2-thinking-20251106
- Creator
- moonshotai
- Added to OpenRouter
- 06.11.2025
- Context length
- 262,144
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.600 / out $2.5
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | No measurement | reasoning: mandatory · parameters: include_reasoning, reasoning |
| Coding | CC — below the B band | — |
| Agents | No measurement | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: mandatory · parameters: include_reasoning, reasoning |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Coding index | 21 | CC — below the B band | 13.11.2026 |
| Design Arena | models · websitewin rate 48.8% · rank 111 | 1,106Elo | CC — below the B band | 13.11.2026 |
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Novitabf16 | 262,144 | $0.600 | $2.5 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.