Kimi K3.1 Pro

Moonshot AI's open-weight 1M-context flagship for long-horizon coding

Context window

Input / 1M tokens

$2.00

Output / 1M tokens

$10.00

Provider

Moonshot AI

Cached input $0.20 per 1M; cache writes $2.00 (5-minute TTL) or $4.00 (1-hour TTL) per 1M. Kimi K3.1 Pro Highspeed (kimi-k3.1-pro-highspeed) runs 2-3x faster for $4.00/$20.00 per 1M. Moonshot says full weights will be released by November 4, 2026; the license has not been published. · Data verified 2026-10-25

Kimi K3.1 Pro (October 23, 2026) is a post-trained version of Kimi K3 from Chinese lab Moonshot AI, with substantially more post-training compute. It has a 1M-token context window, always-on reasoning with low, high and max effort levels, and image and video input. It costs $2/$10 per million tokens, a third less than Kimi K3, and Artificial Analysis lists 2.8 trillion total parameters and an Intelligence Index score of 55. Moonshot says it removed cyber-related data from training before the open release, and full weights are due by November 4.

How to access it

Available on Kimi.com, Kimi Work, Kimi Code, the Kimi mobile apps and the Kimi API (model IDs kimi-k3.1-pro and kimi-k3.1-pro-highspeed) through Chat Completions, Responses and Messages endpoints. Full weights are due by November 4, 2026.

Get access →Documentation →

Strengths

✓1M-token context window for long-horizon coding and agent work
✓Lower price than Kimi K3 ($2.00/$10.00 vs $3.00/$15.00 per 1M tokens)
✓65.7% on Terminal-Bench 4.0 (Moonshot-reported)
✓Open weights promised by November 4, 2026
✓Fewer false claims than K3 on Moonshot's internal stress test (8.3% vs 23.4%)

Best for developers who...

Long-horizon coding agentsFinancial research and document-heavy knowledge work3D and CAD scripting tasksTeams planning to self-host an open-weight frontier-class model

When to choose it (and when not to)

Reach for Kimi K3.1 Pro when...

→You want a frontier-class coding model at $2/$10 per 1M tokens
→You plan to self-host once the weights are released
→You need a 1M-token context for large codebases or document sets

Look elsewhere if...

✕You need the weights or a published license today
✕You need strong offensive-security capability (Moonshot removed cyber data on purpose)

How to use it

›Reasoning is always on; set the effort level (low, high or max) to trade speed for depth
›Use the Highspeed variant when latency matters more than cost

Quickstart

Python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MOONSHOT_API_KEY"],
    base_url="https://api.moonshot.cn/v1",
)

completion = client.chat.completions.create(
    model="kimi-k3.1-pro",
    messages=[{"role": "user", "content": "Your prompt here"}],
)
print(completion.choices[0].message.content)

Base URL from the Kimi K3.1 Pro quickstart

API model id: kimi-k3.1-pro

Benchmarks

Benchmark	Score	Notes
Terminal-Bench 4.0	65.7%	Moonshot-reported; Claude Opus 5.5 66.4%, GPT-6 Astra 58.2%
BenchCAD (with tools)	92.7%	Moonshot-reported
OfficeQA Pro v2	64.4	Moonshot-reported
Artificial Analysis Intelligence Index	55	Third-party; ranked 4th

Source: Kimi K3.1 Pro tech blog

Compare Kimi K3.1 Pro

Compare Kimi K3.1 Pro with any other model

Build a comparison →All model comparisons →

Learn the concepts

MoE (Mixture of Experts)Context Window Inference

← All AI models