Updated 24 October 2026 · Independent resource
Kimi K3.1 Pro
Kimi K3.1 Pro is Moonshot AI's 3T-parameter open-weight language model, released on 23 October 2026. It costs $2 per million input tokens and $10 per million output tokens — about 33% cheaper than Kimi K3 — and ships with a 1M-token context window. On the Artificial Analysis leaderboard it currently ranks #4.
Price and specification figures below are taken from the official Kimi release notes published 23 October 2026, cross-checked against Artificial Analysis and OpenRouter the following day. Last verified: 24 October 2026.
Kimi K3.1 Pro pricing at a glance
Kimi K3.1 Pro uses straightforward pay-per-token pricing with no minimum commitment: you are billed per million tokens processed, split into input (what you send) and output (what the model writes). The headline rate is unchanged for both the standard and long-context paths. The full Kimi K3.1 Pro pricing breakdown — Highspeed rate, cached-input discount, provider comparison and a token cost calculator — is on its own page.
| Model | Input (per 1M) | Output (per 1M) | Change vs K3 |
|---|---|---|---|
| Kimi K3.1 Pro | $2.00 | $10.00 | −33% |
| Kimi K3 | $3.00 | $15.00 | baseline |
A Kimi K3.1 Pro Highspeed variant is also offered for lower-latency serving. It is priced separately from the standard rate — see the full Kimi K3.1 Pro pricing breakdown for the provider-by-provider comparison and a token cost calculator.
Kimi K3.1 Pro benchmarks and specifications
The model is a direct successor to Kimi K3 rather than a new architecture. The headline improvements reported at launch are concentrated in reliability and long-context behaviour, not raw parameter count.
| Specification | Kimi K3.1 Pro | Notes |
|---|---|---|
| Release date | 2026-10-23 | Announced via official blog and social channels |
| Total parameters | 3T | Open-weight model |
| Context window | 1,000,000 tokens | Long-context ceiling unchanged from K3 |
| Artificial Analysis rank | #4 | Composite leaderboard position at launch |
| Hallucination rate | −65% | Reduction reported versus K3 |
| Cyber capability | Reduced | Deliberately limited at launch as a safety trade-off |
Read the cyber note carefully. K3.1 Pro was shipped with a deliberately reduced offensive-security capability. If your workload involves penetration testing, exploit generation or security research, benchmark it against your actual tasks before migrating — the general-purpose leaderboard position does not describe this dimension.
How Kimi K3.1 Pro compares with Kimi K3
The practical summary for anyone deciding whether to migrate: identical context ceiling, one-third lower price on both input and output, a large drop in hallucination rate, and a narrower security surface. For most chat, extraction, summarisation and coding-assistant workloads the price cut alone justifies the switch; for security-adjacent work it does not.
What this site covers
Official rates, the Highspeed variant, third-party provider comparison and a live token cost calculator.
Kimi K3.1 Pro pricing →Leaderboard positions, per-capability results and how K3.1 Pro stacks up against K3 and competing models.
Kimi K3.1 Pro benchmarks →What actually changed in the 23 October release, with links to the primary sources.
Kimi K3.1 Pro release notes →Frequently asked questions about Kimi K3.1 Pro
How much does Kimi K3.1 Pro cost?
$2 per million input tokens and $10 per million output tokens on the official platform — 33% below Kimi K3, which was $3 and $15. Third-party inference providers may quote different rates.
When was Kimi K3.1 Pro released?
23 October 2026. The announcement went out through the official Kimi blog and Moonshot AI's social channels, and the model appeared on third-party platforms the same day.
What is the context window of Kimi K3.1 Pro?
1,000,000 tokens, unchanged from K3. That is roughly 700,000–750,000 English words in a single request, though effective recall at the extreme end of the window is always worth testing on your own data.
Is Kimi K3.1 Pro open source?
It is released as an open-weight model, so the weights can be self-hosted. Check the licence terms in the official release notes before commercial deployment — "open weights" and "open source" are not the same thing.
Where can I use Kimi K3.1 Pro via API?
Through the official Kimi API platform, or through third-party providers such as OpenRouter. Compare per-million-token prices before committing, since hosted rates and rate limits vary considerably.