Kimi K3.1 Pro
Kimi K3.1 Pro is Moonshot's first K3.1 model, announced October 24, 2026. The API ID is kimi-k3.1-pro.
Provider
Moonshot AI
Model family
Moonshot Kimi
Frontier LLM
Cost tier
K3 1 Pro
Status
Current
Release Oct 24, 2026
Why teams choose it
Use kimi-k3.1-pro-highspeed when you need the same model at about 2–3× output speed and
the higher Highspeed token price.
K3-series cache writes are billed separately
Cached input is the cache-hit price only.
K2-series models stop being offered to newly registered users on October 31
2026, and Moonshot plans to retire them by the end of December 2026.
Tradeoffs to know
- Full weights were not released on the pages checked October 30, 2026. Moonshot says they will be released by November 4, 2026.
- Kimi K3 remains a separate API model. K3.1 Pro does not replace that ID.
- Benchmark scores are Moonshot's reported figures. The English API guide's score grid was still a placeholder.
Benchmarks
- Source (vendor-reported)
- https://www.kimi.com/news/kimi-k3-1-pro
- Officeqa Pro V2 (vendor-reported)
- 64.4
- Terminal Bench 4 0 (vendor-reported)
- 0.657
- Benchcad With Tools (vendor-reported)
- 0.927
- Kimi Code Bench 3 Long Horizon (vendor-reported)
- 0.612
Technical specs
- Inputs
- text, image
- Outputs
- text
- License
- Proprietary API
- API ID
kimi-k3.1-pro- Context tokens
- 1,048,576
- Input Usd Per Million
- 2
- Output Usd Per Million
- 10
- Cache Read Usd Per Million
- 0.2
- Cache Write 1h Usd Per Million
- 4
- Cache Write 5m Usd Per Million
- 2
- Capabilities
- 1,048,576-token context window
- Vision input, tool calling, and coding agents including Kimi Code
- Flat pay-as-you-go pricing with 5-minute and 1-hour cache-write rates
Moonshot Kimi family lineup
Related models
Kimi K3.1 Pro FAQ
What is Kimi K3.1 Pro?
1 model, announced October 24, 2026. 1-pro. Moonshot describes it as a post-trained Kimi K3, a 3-trillion-class model with a 1,048,576-token context window and native visual understanding. com, the Kimi app, Kimi Work, Kimi Code, and the Kimi API. 00 for 1-hour cache writes, per million tokens. Cache billing is automatic, with a default 5-minute TTL. Moonshot says the full weights will be released by November 4, 2026. 4. The English API guide's benchmark grid still showed placeholders on October 30, 2026.
When does Kimi K3.1 Pro fit best?
Long-horizon coding in Kimi Code
What should teams watch out for with Kimi K3.1 Pro?
Full weights were not released on the pages checked October 30, 2026. Moonshot says they will be released by November 4, 2026.
Explore next
Models, tools, and comparisons that connect to this reference.