Moonshot's Kimi K3.1 Pro Debuts, Weights Due Nov 4
The reasoning-focused model is live via API now, with Moonshot promising full open weights in early November.
Moonshot AI has introduced Kimi K3.1 Pro, a text and reasoning model that is available first through the company's API, with full open weights promised by November 4, 2026. The staggered rollout — API access now, downloadable weights later — gives developers an early look while the release is formalized (announcement details).
The model lands in the LLM-text category with an emphasis on reasoning, continuing Moonshot's push to position Kimi as a serious contender among open-leaning labs. The company has not yet published parameter counts, context-window figures, or benchmark results, and the license is listed only as "other" pending the weights drop.
Why it matters
The gap between an API launch and an open-weights release is where a lot of the real-world story plays out. A few things worth watching:
- Whether the promised November 4 weights arrive on schedule, and under what license terms.
- How K3.1 Pro's reasoning performance compares to other recent open releases once independent testing is possible.
- What context length and architecture Moonshot ultimately confirms.
For now, teams curious about Kimi can evaluate the model through the API while waiting for the open release. The authoritative blog post is available at kimi.ai, and we'll update this story as specs and licensing details are confirmed.
Sources
- Visit
Introducing Kimi K3.1 Pro
Announcement
More from Moonshot AI
All Moonshot AI releases →
Moonshot AI releases Kimi K3, a 2.8T-parameter MoE model
The open-weights multimodal model leans into coding and agentic tasks, extending Moonshot's Kimi line into a new scale bracket.
Moonshot AI Releases Kimi, a Multimodal Coding Model
The new Mixture-of-Experts model from the Chinese AI company can generate code while also understanding visual inputs, a rare combination in open models.
Kimi K2.6 tops closed models in coding test
Moonshot AI's open-weights mixture-of-experts model reportedly outperformed Claude, GPT-5.5, and Gemini on a programming challenge.
More in Text / LLM
All Text / LLM →
Recursive releases Flint, an Apache-2.0 chat model
The new conversational LLM ships under a permissive license with community GGUF builds already available for local use.
KRAFTON releases Raon-1-mini preview under Apache 2.0
The Korean game studio behind PUBG puts out a compact, permissively licensed language model as an early preview.
DeepSeek V4.1-Flash Adds Vision and Reasoning
The new MIT-licensed checkpoint folds image understanding and reasoning into DeepSeek's fast, mixture-of-experts Flash line.
0 comments
No comments yet. Be the first to weigh in.