≫ Home / LLM API pricing / Kimi K3

Kimi K3 API pricing

Of the 7 providers checked, SiliconFlow lists the lowest blended rate for Kimi K3: $2.70 input and $13.50 output per 1M tokens, on its pricing page as of 25 September 2026.

Moonshot AI's open-weight model, reading text, images and video with a 1M-token context. At 2.8T parameters it is the largest model on this list, though each token uses only 16 of its 896 experts. The weights use Moonshot's own Kimi K3 License.

Creator
Moonshot AI
Weights
Open
License
Kimi K3 License
Size
2.8T total, 104B active
Context
1M tokens
Input
text, image, video
Model card
Hugging Face

Facts from the model card, checked 25 Sep 2026.

Across 7 providers, the most expensive charges 1.3× the cheapest on a blended basis.

Prices by provider

Per 1M tokens, on-demand, sorted cheapest first by blended cost (3:1 input to output). Off-peak and long-context rates are listed on each provider's page.

# Provider Input / 1M Cached / 1M Output / 1M Checked Notes
1 SiliconFlow $2.70 $0.27 $13.50 25 Sep 2026
2 Aurora Inference $2.85 $0.29 $14.25 25 Sep 2026
3 DeepInfra $2.85 $0.285 $14.25 25 Sep 2026
4 Baseten $3.00 $0.30 $15.00 25 Sep 2026
5 Fireworks AI $3.00 $0.30 $15.00 25 Sep 2026 Standard tier
6 Together AI $3.00 $0.30 $15.00 25 Sep 2026
7 Berget AI €3.00 ≈ $3.4101 €3.00 €15.00 ≈ $17.0505 27 Sep 2026 Cached input is billed at the input price

Each price was read off the provider's own pricing page on the date shown; the date links to that page. Providers that publish no per-token price for this model are not listed. EUR prices are ranked at the ECB reference rate of 24 September 2026. The same data is open under CC-BY at /api/query/prices.

Compare every inference provider on free tiers, hosting and EU availability in the inference APIs table.

Is your product missing?

Add it here →