Moonshot AIopen weightsKimi K2.6 pricing
$0.65 per 1M input tokens and $3.41 per 1M output tokens. In real work that is 0.33¢ per chat answer and 1.1¢ to summarize a 20-page PDF.
Official price at Moonshot (Kimi): $0.95 / $4.00 per 1M, cached $0.16 (source, verified Oct 18, 2026). OpenRouter lists $0.65 / $3.41, which is the cheapest route across its hosts - lower than the official price.
What Kimi K2.6 costs per task
Token counts estimated from a measured model of the same lab (MiniMax M3), rescaled by tokenizer. Reasoning tokens are billed as output.
| Task | Cost | Tokens used | Price rank | Basis |
|---|---|---|---|---|
| Answer a chat question | 0.33¢$3.33 / 1,000 | 208 in · 454 out · 482 reasoning | 203 of 336 | estimated |
| Reply to a customer email | 0.18¢$1.81 / 1,000 | 424 in · 264 out · 187 reasoning | 223 of 336 | estimated |
| Write a 1,500-word blog post | 0.78¢$7.80 / 1,000 | 222 in · 1,804 out · 442 reasoning | 165 of 336 | estimated |
| Write 10 product descriptions | 0.58¢$5.82 / 1,000 | 448 in · 1,089 out · 531 reasoning | 168 of 336 | estimated |
| Translate 1,000 words | 2.1¢$20.62 / 1,000 | 1,498 in · 1,227 out · 4,535 reasoning | 246 of 336 | estimated |
| Proofread one page (500 words) | 0.43¢$4.25 / 1,000 | 888 in · 662 out · 416 reasoning | 218 of 336 | estimated |
| Summarize a 20-page PDF | 1.1¢$10.97 / 1,000 | 13,121 in · 469 out · 246 reasoning | 195 of 336 | estimated |
| Extract data from an invoice | 0.33¢$3.29 / 1,000 | 545 in · 572 out · 289 reasoning | 199 of 336 | estimated |
| Analyze a 200-row spreadsheet | ≥5.8¢$57.83 / 1,000 | 5,024 in · 155 out · 15,847 reasoning | 249 of 336 | estimated |
| Answer from 5 retrieved documents (RAG) | 0.68¢$6.83 / 1,000 | 3,837 in · 836 out · 437 reasoning | 203 of 336 | estimated |
| Classify 100 support tickets | 0.75¢$7.53 / 1,000 | 1,448 in · 448 out · 1,485 reasoning | 182 of 336 | estimated |
| Review 500 lines of code | 1.2¢$12.12 / 1,000 | 4,843 in · 2,632 out | 167 of 336 | estimated |
| Write a SQL query | 0.26¢$2.61 / 1,000 | 276 in · 714 out | 182 of 336 | estimated |
| Run a 50-step coding agent | 46¢$464 / 1,000 | 2,259,875 in · 15,150 out · 9,350 reasoning | 163 of 336 | modeled |
Price at every provider
List prices per 1M tokens (input / output) at each host serving Kimi K2.6 through OpenRouter today. Hosts can differ in quantization, context and speed.
| Provider | Input / output | Cached input | Context | Uptime |
|---|---|---|---|---|
| B Baidufp4 | $0.43 / $1.82 | $0.073 | 262K | 100% |
 Chutesint4 | $0.45 / $1.95 | $0.045 | 262K | |
| D DigitalOceanunknown | $0.57 / $2.40 | $0.11 | 262K | 100% |
| S StreamLakefp8 | $0.60 / $2.52 | $0.10 | 256K | 100% |
| I Inceptronint4 | $0.43 / $3.20 | $0.16 | 262K | 100% |
| C CoreWeavefp4 | $0.65 / $3.41 | $0.15 | 262K | 100% |
 SiliconFlowfp8 | $0.77 / $3.40 | $0.14 | 262K | |
 Novitaunknown | $0.80 / $3.40 | $0.16 | 262K | 100% |
| C Crusoebf16 | $0.70 / $3.50 | $0.35 | 262K | 100% |
 DeepInfrafp4 | $0.75 / $3.50 | $0.15 | 262K | 100% |
| V Veniceint4 | $0.75 / $3.50 | $0.16 | 256K | 100% |
 Parasailint4 | $0.75 / $3.50 | $0.16 | 262K | 100% |
| G GMICloudfp8 | $0.85 / $3.60 | $0.14 | 262K | |
| A AtlasCloudint4 | $0.95 / $4.00 | $0.16 | 262K | 100% |
| C Cloudflareunknown | $0.95 / $4.00 | $0.16 | 262K | |
Moonshot AIint4 | $0.95 / $4.00 | $0.16 | 262K | 100% |
| P Phalaunknown | $1.09 / $4.60 | $0.37 | 262K |
Price history
First seen in our price history on Apr 23, 2026 at $0.56 input / $3.50 output. Since then the output price moved from $3.50 to $3.41 per 1M tokens (18 changes).
FAQ
How much does Kimi K2.6 cost?
Kimi K2.6 costs $0.65 per 1 million input tokens and $3.41 per 1 million output tokens, with cached input at $0.15 (list price on Oct 18, 2026). In practice one chat answer costs about 0.33¢ and summarizing a 20-page PDF about 1.1¢.
Where is Kimi K2.6 cheapest?
Among 17 providers we track, the lowest list price today is at Baidu: $0.43 input and $1.82 output per 1M tokens.
Is the Kimi K2.6 cost per task measured or estimated?
Estimated: Kimi K2.6 was not run directly. Tokens come from MiniMax M3 and are rescaled by the tokenizer of Moonshot AI. Real usage can differ.