Pricingby GlyphKimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI.
Ways to buy 30
| Way to buy | Input | Output | In / out per 1M | Cached | Vs list |
|---|---|---|---|---|---|
| InferenceNet us, fp4, cheapest · cached $0.21 | $0.239 | $10.8 | $0.239 / $10.8-52% vs list | $0.21 | -52% vs list |
| Sail Research fp4-us, fp4 · cached $0.4 | $1.2 | $9 | $1.2 / $9-47% vs list | $0.4 | -47% vs list |
| Sail Research fp4 · cached $0.33 | $0.885 | $10.5 | $0.885 / $10.5-45% vs list | $0.33 | -45% vs list |
| ParetoAlpha us · cached $0.225 | $1 | $11.3 | $1 / $11.3-41% vs list | $0.225 | -41% vs list |
| Makora fp4 · cached $0.2 | $0.5 | $13 | $0.5 / $13-40% vs list | $0.2 | -40% vs list |
| Morph fp8 · cached $0.218 | $1.5 | $10.5 | $1.5 / $10.5-37% vs list | $0.218 | -37% vs list |
mxfp4 · cached $0.185 | $1.85 | $9.5 | $1.85 / $9.5-37% vs list | $0.185 | -37% vs list |
| InferenceNet fp4 · cached $0.3 | $1.3 | $11.4 | $1.3 / $11.4-36% vs list | $0.3 | -36% vs list |
| Relace fp4 · cached $0.29 | $1.8 | $11.2 | $1.8 / $11.2-31% vs list | $0.29 | -31% vs list |
| Io Net fp4 · cached $0.2 | $1.5 | $14 | $1.5 / $14-23% vs list | $0.2 | -23% vs list |
Default region · cached $0.255 | $2.55 | $13 | $2.55 / $13-14% vs list | $0.255 | -14% vs list |
| Wafer Default region, mxfp4 · cached $0.3 | $1.99 | $15 | $1.99 / $15-13% vs list | $0.3 | -13% vs list |
| InferenceNet us-fast, fp4 · cached $0.35 | $2.24 | $14.5 | $2.24 / $14.5-12% vs list | $0.35 | -12% vs list |
mxfp4 · cached $0.285 | $2.85 | $14.3 | $2.85 / $14.3-5% vs list | $0.285 | -5% vs list |
| Wafer fast, mxfp4 · cached $0.4 | $0.5 | $22 | $0.5 / $22-2% vs list | $0.4 | -2% vs list |
| Wafer us, mxfp4 · cached $0.3 | $2.8 | $15.5 | $2.8 / $15.5Same as list | $0.3 | Same as list |
The lab's own API · cached $0.3 | $3 | $15 | $3 / $15List price | $0.3 | List price |
 AWS Bedrockus-east-2 · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
fp8 · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
| Decart mxfp4 · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
Default region · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
mxfp4 · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
fp4 · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
| Phala Default region · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
Default region · cached $0.3 | $3 | $15 | $3 / $15Same as list | $0.3 | Same as list |
| InferenceNet fast, fp4 · cached $0.45 | $3.25 | $16 | $3.25 / $16+7% vs list | $0.45 | +7% vs list |
Default region · cached $0.345 | $3.45 | $17.3 | $3.45 / $17.3+15% vs list | $0.345 | +15% vs list |
fast · cached $0.45 | $4.5 | $22.5 | $4.5 / $22.5+50% vs list | $0.45 | +50% vs list |
us · cached $0.45 | $4.5 | $22.5 | $4.5 / $22.5+50% vs list | $0.45 | +50% vs list |
fast, fp4 · cached $0.45 | $4.5 | $22.5 | $4.5 / $22.5+50% vs list | $0.45 | +50% vs list |
USD per 1M tokens; batch $0.917 in, $6 out on the lab's API. Regions are separate routes.
Through a gateway
Your own numbers100M input and 20M output tokens a month: $600.00 direct.
Kong AI Gateway$600.00+0%
LiteLLM$600.00+0%
Neon AI Gateway$600.00+0%
Vercel AI Gateway$600.00+0%
Cloudflare AI Gateway$630.00+5%
Requesty$630.00+5%
Eden AI$633.00+5.5%
OpenRouter$633.00+5.5%
HeliconeNot offeredNot offered
TrueFoundry AI GatewayNot offeredNot offered
A dash: not on that gateway's own model list. No published token fee: Martian Gateway, Portkey.
Related models
| In / out per 1M | Cheapest route | |||||||
|---|---|---|---|---|---|---|---|---|
Moonshot AI · 1M | $0.377 | $7.95 | $0.377 / $7.95 | $2.27 | 1M | on Relace (fp4) | · | Kimi K3.1 Pro is Moonshot AI's flagship open-weight multimodal reasoning model and a post-trained successor to Kimi K3, with a 1M-token context window. |
Moonshot AI · 262KCheapest: $1.2 on CoreWeave (int4) | $0.95 | $4 | $0.95 / $4 | $1.71 | 262K | $1.2 on CoreWeave (int4) | · | MoonshotAI: Kimi K2.7 Code is a coding-focused model in Moonshot AI's Kimi K2 family, built to complete end-to-end programming tasks reliably over long contexts. |
Moonshot AI · 262KCheapest: $0.742 on Chutes (int4) | $0.95 | $4 | $0.95 / $4 | $1.71 | 262K | $0.742 on Chutes (int4) | · | Kimi K2.6 is Moonshot AI's next-generation multimodal model, designed for long-horizon coding, coding-driven UI/UX generation, and multi-agent orchestration. |
Moonshot AI · 262K | $0.49 | $2.5 | $0.49 / $2.5 | $0.992 | 262K | on AtlasCloud (int4) | · | Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm. |
OpenAI · 1.05MCheapest: $2 on OpenAI (flex) | $2 | $10 | $2 / $10 | $4 | 1.05M | $2 on OpenAI (flex) | · | GPT-6.1 Sol is an upgrade to GPT-6 Sol from OpenAI, positioned below the flagship GPT-6 Astra in the GPT-6 series. |
Anthropic · 1M | $4 | $20 | $4 / $20 | $8 | 1M | Lab list price | · | Claude Opus 5.5 is Anthropic's flagship model for demanding reasoning, coding, and long-horizon agentic work, succeeding Claude Opus 5. |
Google · 1MCheapest: $2 on Google Gemini API (flex) | $2 | $10 | $2 / $10 | $4 | 1M | $2 on Google Gemini API (flex) | · | Gemini 4 Argon is Google's flagship reasoning model, built for complex software engineering, deep mathematical and scientific problem solving, and autonomous multi-step agents. |
Anthropic · 1M | $2 | $10 | $2 / $10 | $4 | 1M | Lab list price | · | Claude Sonnet 5.5 is Anthropic's Sonnet-class model for well-scoped everyday work, succeeding Claude Sonnet 5 as a direct upgrade. |
About Kimi K3
Takes text, image, video in and returns text. First listed Sep 23, 2026. Up to 944K output tokens. Blended price $6 per 1M tokens at 3 input to 1 output. Overall score 4 on leadersboard.fru.dev (rank 29). Model id kimi-k3.
Price history
List price since Oct 27, 2026, from daily reads of the OpenRouter catalog.
Kimi K3
Io Net (fp4) cut its price: input $2.3 to $1.5 per 1M tokens.
ParetoAlpha (us) cut its price: input $3 to $1 and output $15 to $11.3 per 1M tokens.
AkashML (fp4) cut its price: output $15 to $14 per 1M tokens.
Makora (fp4) cut its price: output $15 to $13 per 1M tokens.
Makora (fp4) cut its price: input $0.65 to $0.5 per 1M tokens.
InferenceNet (us) cut its price: output $11.9 to $10.8 per 1M tokens.
AkashML (fp4) cut its price: input $3 to $1.2 per 1M tokens.
Parasail (fast) raised its price: input $3 to $4.5 and output $15 to $22.5 per 1M tokens.
Makora (fp4) cut its price: input $1.53 to $0.65 and output $12.8 to $15 per 1M tokens.
Questions
How much does Kimi K3 cost?
$3 per million input tokens and $15 per million output tokens, with cached input at $0.3, or $0.917 and $6 through the Batch API.
What is the context window of Kimi K3?
1M tokens, with up to 944K tokens of output.
Where is Kimi K3 cheapest?
Of 30 routes tracked, InferenceNet is cheapest at $0.239 input and $10.8 output per million tokens.
Has Kimi K3's price changed?
Yes. The input price went from $2.3 to $1.5, first seen Nov 8, 2026.