Post-Cutoff
Search For AI agents

Moonshot AIKimi K3 familyCurrentReasoning modelReleased 23 Oct 2026

Kimi K3.1 Pro

Moonshot’s follow-up to Kimi K3. API live at launch on Oct 23, 2026; weights promised by Nov 4, 2026.

Model idkimi-k3.1-pro

Price$2 in, $10 out per 1M tokens

Checked24 October 2026

No published training cutoff

At least5events since its release on 23 October 2026

0 historic, 0 major. 2 days of news.

Moonshot AI doesn’t publish a training cutoff, so we count from its release date. The real gap is larger.

How to call it

4 ways to use Kimi K3.1 Pro, with the model id each provider expects.

Providers, model ids, endpoints and docs for Kimi K3.1 Pro
ProviderModel idEndpointDocs
Kimi API (Moonshot) kimi-k3.1-pro https://api.moonshot.ai/v1 platform.kimi.ai
Kimi API (Moonshot), fast variant kimi-k3.1-pro-highspeed https://api.moonshot.ai/v1
OpenRouter moonshotai/kimi-k3.1-pro openrouter.ai
Web app kimi.com

Pricing

Prices for Kimi K3.1 Pro
Input$2 per 1M tokens
Output$10 per 1M tokens
Cache read$0.20 per 1M tokens

As published: “per 1M tokens (USD)”.

Source: openrouter.ai, checked 24 October 2026.

Specs

Input
Text and images
Output
Text
Context window
1,048,576 tokens
Max output
Not published
Open weights
No
Licence
proprietary until weights are released (promised by 2026-11-04; licence not yet named)
Training cutoff
Not published
Released
23 October 2026
Status
Current
Type
Reasoning model

What stands out

What its maker marketed at launch, and what people found later, each with its source.

  • Near-frontier score for a soon-open model

    First

    Artificial Analysis Intelligence Index 55 at launch, 4th overall behind Claude Opus 5.5 (58), Sonnet 5.5 (56) and Muse Wonder Full (56); highest Terminal-Bench 4.0 score AA had measured (67%); GDPval-AA 1,846 Elo.

    Source: x.com

  • Cyber-capability data removal

    First

    Moonshot removed cyber-related data from both mid-training and RL so the open weights would not extend the frontier of offensive cyber capability; the unfiltered checkpoint scored 90.5 vs 81.0 on xdev-bench-hard.

    Source: kimi.ai

  • Game-engine and CAD work

    Marketed for full-project coding in Unreal, Unity and Godot and for parametric 3D modelling in Blender and CAD (BenchCAD with tools 92.7% Voxel IoU).

    Source: kimi.ai

Notes

2.8T total / 104B active MoE, 1M context. Highspeed variant is 2–3x faster at about twice the price, with the same cached-input price (Simon Willison). Defaults: reasoning effort max, temperature 1.0, top_p 0.95. Enterprise admins must enable it (off by default). Zero Data Retention available on the API.

curl https://api.moonshot.ai/v1/chat/completions \
 -H "Authorization: Bearer $MOONSHOT_API_KEY" -H "Content-Type: application/json" \
 -d '{"model":"kimi-k3.1-pro","reasoning_effort":"max","messages":[{"role":"user","content":"Hello"}]}'

Sources: https://www.kimi.ai/blog/kimi-k3-1-pro · https://openrouter.ai/moonshotai/kimi-k3.1-pro · https://simonwillison.net/2026/Oct/23/kimi-k31-pro/

Not yet checked against an official source: Max output and knowledge cutoff not verified; Kimi’s own pricing page did not render prices for our fetcher, so prices are OpenRouter’s (matching Willison’s $2/$10).