Kimi K2 Thinking.
Moonshot's open-weight reasoning variant — extended chain-of-thought training on top of Kimi K2.
4× AMD MI300.
Most-aggressive quantisation we have a working recommendation for. Lower precision = less VRAM = cheaper hardware, at a small accuracy cost.
Cheapest hosted endpoints.
Speed across providers.
Tokens/sec and time-to-first-token measured against the same prompt template on each provider's API.
| Provider | Tokens/sec | TTFT | Total |
|---|---|---|---|
| OpenRouter | 49.8 | 25704 ms | 27832 ms |
Variants in the Kimi family.
Frequently asked.
How do I run Kimi K2 Thinking?
Where can I access Kimi K2 Thinking?
How much does it cost to run Kimi K2 Thinking?
Is Kimi K2 Thinking open-source or proprietary?
Cheapest hardware per quantisation.
Each row is one quantisation tier (the same weights compressed differently). Lower precision → lower VRAM → cheaper hardware, at the cost of small accuracy loss. $/hr refreshed hourly from each provider's API.
| Quantisation | Cheapest GPU config | Total VRAM | Live $/hr | tokens/sec | |
|---|---|---|---|---|---|
|
FP8
FP8 — 8-bit float (Hopper / Blackwell)
|
1536 GB | — | — | Compare → | |
|
INT4
INT4 — 4-bit integer (~4× VRAM saving)
|
768 GB | — | — | Compare → |
What it costs per month across providers.
Estimate your monthly bill for Kimi K2 Thinking across every host that publishes per-token pricing. Slide your token volumes; the chart + table re-rank cheapest-first.
Cheapest provider on the left.
Total monthly cost — input + output tokens combined.
Bill breakdown.
About Kimi K2 Thinking.
Kimi K2 Thinking is Moonshot AI's reasoning-specialised variant of Kimi K2. Same Mixture-of-Experts base (1T parameters total, 32B active per token) but post-trained with extended thinking-token allocation — the model emits a long visible reasoning trace before its final answer, similar to DeepSeek R1's pattern. Wins on math and complex multi-step coding benchmarks where standard K2 underperforms. Available on Hugging Face under the same modified MIT licence as the base K2. Accessed via Moonshot's platform, Kimi.com (consumer), Ollama, and self-hosted on rented multi-GPU clusters.