Kimi K2.7 Code Highspeed API pricing
Official Kimi list price, last checked Oct 29, 2026. Updated Oct 29, 2026.
Kimi K2.7 Code Highspeed costs $1.90 per 1M input tokens and $8 per 1M output tokens, cached input at $0.38, and a 262K-token context window, per Kimi's official pricing page (checked Oct 29, 2026). It's a premium-priced model: number 59 of 77 by blended price ($3.42 per 1M tokens at a 3 to 1 input to output mix). LMArena doesn't list it on the boards we track, so we don't rank its quality. For less, look at GPT-6.1 Terra ($2 blended).
Kimi K2.7 Code Highspeed price per million tokens
| Kimi K2.7 Code Highspeed price per million tokens | USD |
|---|---|
| Input | $1.90 |
| Cached input | $0.38 |
| Output | $8 |
| Blended, 3 parts input to 1 part output | $3.42 |
| Context window | 262K tokens |
Official note from Kimi: About 180 tokens per second. Closed to new sign-ups from Oct 31, 2026. Moonshot plans to retire the K2 series by the end of December 2026.
What Kimi K2.7 Code Highspeed costs for real work
List-price estimates for a month, no discounts. Our math, not a quote: change the token counts in the plan vs API calculator.
| Workload | Tokens a month | Kimi K2.7 Code Highspeed | Qwen3.7 Max | GPT-6.1 Terra |
|---|---|---|---|---|
| Light: a side project or personal assistant | 2M in, 0.5M out | $7.80 | $8.75 | $4.50 |
| Medium: a small team or one product feature | 20M in, 5M out | $78.00 | $87.50 | $45.00 |
| Heavy: a production app | 200M in, 50M out | $780 | $875 | $450 |
| Medium with 80% of input cached | 20M in (16M cached), 5M out | $53.68 | n/a | $29.80 |
How Kimi K2.7 Code Highspeed ranks on LMArena
It isn't on any LMArena board we track as of Oct 29, 2026. We don't estimate quality for unranked models. See the best model for each task for ranked options.
Closest rivals to Kimi K2.7 Code Highspeed
The models nearest in price, other makers first. Ranks are on LMArena, where we have them.
| Model | In / out | Blended | Verdict |
|---|---|---|---|
| Qwen3.7 Max | $2.50 / $7.50 | $3.75 | Kimi K2.7 Code Highspeed wins on price. Compare |
| Qwen3.8 Max | $2 / $6 | $3 | Qwen3.8 Max wins on price. Compare |
| Qwen3.8 2.4T-A95B | $2 / $6 | $3 | Qwen3.8 2.4T-A95B wins on price. Compare |
| Mistral Medium 3.5 | $1.50 / $7.50 | $3 | Mistral Medium 3.5 wins on price. Compare |
Cheaper alternatives to Kimi K2.7 Code Highspeed
At least 40% cheaper blended, cheapest at the bottom of the list we found.
| Model | In / out | Blended | Verdict |
|---|---|---|---|
| GPT-6.1 Terra | $1 / $5 | $2 | GPT-6.1 Terra wins on price. Compare |
| Muse Spark 1.3 | $1.25 / $4.25 | $2 | Muse Spark 1.3 wins on price. Compare |
| DeepSeek V4 Pro | $1.32 / $3.96 | $1.98 | DeepSeek V4 Pro wins on price. Compare |
More from Kimi
- Kimi K3.1 Pro$2 in, $10 out per 1M
- Kimi K2.7 Code$0.95 in, $4 out per 1M
- Kimi K2.6$0.95 in, $4 out per 1M
- Kimi K3$3 in, $15 out per 1M
- Kimi K3.1 Pro Highspeed$4 in, $20 out per 1M
FAQ
How much does the Kimi K2.7 Code Highspeed API cost per million tokens?
$1.90 per 1M input tokens and $8 per 1M output tokens, plus $0.38 per 1M for cached input, as of Oct 29, 2026. A typical 3 to 1 input to output mix works out to $3.42 per 1M tokens.
Does Kimi K2.7 Code Highspeed charge more for long prompts?
Kimi doesn't list a long-prompt surcharge for Kimi K2.7 Code Highspeed. Its pricing note says: About 180 tokens per second. Closed to new sign-ups from Oct 31, 2026. Moonshot plans to retire the K2 series by the end of December 2026.
Is Kimi K2.7 Code Highspeed cheaper than Qwen3.7 Max?
Yes. A medium month (20M input and 5M output tokens) costs $78.00 on Kimi K2.7 Code Highspeed and $87.50 on Qwen3.7 Max at list price.
How does Kimi K2.7 Code Highspeed rank on LMArena?
It isn't listed on the LMArena boards we track, so we show no quality rank rather than guess.
What is a cheaper alternative to Kimi K2.7 Code Highspeed?
GPT-6.1 Terra costs $1 in and $5 out per 1M tokens, about 1.7x less blended.
Where does this Kimi K2.7 Code Highspeed price come from?
From Kimi's official pricing page, which we check every day. This page was last checked Oct 29, 2026.