Updated

A brushed-silver crescent sculpture circled by two thin silver rings over dark navy mist on a dark background, abstract hero artwork for Kimi K3.1 Pro

Moonshot AI

Kimi K3.1 Pro

Moonshot AI's Kimi K3 with more post-training, built for long coding sessions, 3D work and finance.

Chat with Kimi K3.1 Pro in Super AgentView pricing

What is Kimi K3.1 Pro?

Kimi K3.1 Pro is the first model in Moonshot AI's K3.1 series, released on October 23, 2026.

Moonshot describes it as a post-trained version of Kimi K3 with substantially larger post-training compute. It has 2.8 trillion parameters (104B active), takes text, image and video input, and has a 1M-token context window. Moonshot says it always thinks, with low, high and max effort. Moonshot says it will release the weights by November 4, 2026.

Provider
Moonshot AI
Context window
1M tokens
Input
Text, Image, Video
Reasoning
Off, low, medium, high, extra high, max
Max output
1M tokens
Released
Oct 23, 2026

How Kimi K3.1 Pro performs

Published by Moonshot AI

Results Moonshot AI reported for Kimi K3.1 Pro. Each lab runs its own setup, so compare these with other models only inside the same table.

Benchmark results Moonshot AI published for Kimi K3.1 Pro
BenchmarkScoreSetting
Terminal-Bench 4.065.7%max effort, Claude Code harness
SWE-Marathon v1.151.7%max effort, Claude Code harness
DeepSWE v1.175.5%max effort, Claude Code harness
BenchCAD (with tools)92.7%max effort, Claude Code harness, 1,000-case subset
Humanity's Last Exam (full)70.0%with tools, max effort
GPQA Diamond94.2%max effort
BrowseComp92.1%max effort, context compaction at 200K, up to 500 steps
OSWorld V2.179.6%max effort, Kimi Agent harness
OfficeQA Pro v264.4max effort, Claude Code harness, PDFs as images
MMMU-Pro84.1%no tools, max effort

Source: kimi.ai

What you can do with Kimi K3.1 Pro

  • Built on Kimi K3 with much more post-training compute, per Moonshot

  • Text, image and video input with a 1M-token context window

  • Long-horizon coding, from quick fixes to full-project builds

  • 3D creation from a text brief, sketch or reference image

  • Finance workflows such as models, decks and cited sector reports

How to prompt Kimi K3.1 Pro

  • Start a new chat with Kimi K3.1 Pro instead of switching to it mid-conversation, as Moonshot advises.
  • For 3D work, give it a text brief, a sketch or a reference image.
  • If answers take too long, try a lower effort; Moonshot suggests low.

Where Kimi K3.1 Pro fits best

  • Long coding sessions in real codebases

  • 3D models and CAD-style parts

  • Financial models, decks and sector reports

  • Tasks that mix text, images and video

How Kimi K3.1 Pro compares

Kimi K3.1 Pro compared with other chat models on Flashloop
ModelLMArena textLMArena codingContext windowReasoning
Kimi K3.1 ProNoNo1M tokensOff, low, medium, high, extra high, max
GPT-6.1 Sol1480#271537#141.05M tokensLow to max
GPT-6.1 Sol ProNoNo1.05M tokensLow to max
Claude Sonnet 5.51478#291528#261M tokensLow to max
GPT-6 Luna1442#911507#651.05M tokensOff to max

Good to know before you start

  • Moonshot says it always thinks and documents only low, high and max effort
  • Moonshot removed cyber data from training, and says its cyber skills trail frontier proprietary models
  • On Moonshot's own table it trails other models on some rows, including Humanity's Last Exam without tools and Program Bench
  • Moonshot's lower hallucination rate comes from an internal stress test and is not an everyday error rate
  • Moonshot has not published the weights, license or knowledge cutoff yet

Start chatting with Kimi K3.1 Pro

Chat with Kimi K3.1 Pro in Super Agent

  1. Select Kimi K3.1 Pro

    Open Super Agent on Flashloop and choose Kimi K3.1 Pro in the model picker.

  2. Enter your prompt

    Ask for what you need: an answer, a plan, or the images and videos you want the agent to make.

  3. Send and keep going

    Send the message. Kimi K3.1 Pro answers in Super Agent and runs any generations you asked for.

Questions about Kimi K3.1 Pro

Is Kimi K3.1 Pro open-weight?

Not yet. Moonshot says it will release the full weights by November 4, 2026.

What inputs does Kimi K3.1 Pro accept?

Text, images and video, with a 1M-token context window.

What reasoning levels does Kimi K3.1 Pro support?

Moonshot documents low, high and max effort and says K3.1 Pro always thinks. Super Agent's picker lists off, low, medium, high, extra high and max.

How is Kimi K3.1 Pro different from Kimi K3?

Moonshot says it is Kimi K3 with much more post-training, on the same 2.8T-parameter base, 1M-token window and text, image and video input. Moonshot reports higher scores on its comparison rows, such as Humanity's Last Exam with tools at 70.0% against 56.0% for K3. These are Moonshot's own figures.

More chat models on Flashloop

Browse all models