Skip to content

Moonshot AI News: Latest Releases & Events

Moonshot AI open-sourced Kimi K3.1 Pro in November 2026, a 2.8T-parameter multimodal model with a 1M-token context window. In October and September, Anthropic and OpenAI alleged Moonshot-linked distillation of their models; in July, Kimi K3 ranked third on the Artificial Analysis Intelligence Index. In February, Anthropic had accused Moonshot, DeepSeek and MiniMax of industrial-scale distillation of Claude.

Who is Moonshot AI? → Big AI Players

All 14 Moonshot AI entries of the AI News timeline, newest first.

2026

November 2026

2026-11-04

  • Moonshot AI Open-Sources Kimi K3.1 Pro
    Moonshot AI released the open weights of Kimi K3.1 Pro, its most capable 3T-class model to date — a natively multimodal Mixture-of-Experts model with 2.8T total parameters (104B activated), a 1-million-token context window, and MXFP4 quantization-aware training. Post-trained from Kimi K3 with large-scale joint reinforcement learning across hundreds of downstream domains, the first K3.1-series model posts strong results across coding, agentic, reasoning, and cybersecurity benchmarks, and is available on Hugging Face and via the Kimi API under the Kimi K3.1 Pro License. (huggingface.co)
    Hugging Face

October 2026

2026-10-25

  • OpenAI, Anthropic slash LLM prices 80%
    An analysis by Max Woolf documents how OpenAI and Anthropic cut LLM prices by up to 80% between June and October 2026 (including an 80% cut to GPT-5.6 Luna, GPT-6 Sol halving prices, and Claude Opus 5.5 reducing total costs 40-60%), with no apparent loss in model quality. The article attributes the price war to tit-for-tat competition, pricing pressure from cheap open-weight rivals like DeepSeek and Kimi, healthy inference margins ahead of rumored IPOs, and improved compute efficiency. (minimaxir.com)
    OpenAI Anthropic DeepSeek

2026-10-23

  • Moonshot AI releases Kimi K3.1 Pro
    Moonshot AI launched Kimi K3.1 Pro, its most capable 3T-class open-weight model (2.8T total / 104B active params MoE, 1M-token context, multimodal), alongside a 2-3× faster K3.1 Pro Highspeed variant, with full weights promised by Nov 4, 2026 — announced by the company on X. The model matches or approaches frontier proprietary models on coding, agentic, and knowledge benchmarks (65.7% on Terminal-Bench 4.0, ahead of GPT-6 Astra, narrowly behind Claude Opus 5.5), and Simon Willison notes it is the first soon-to-be-open-weights model to score this highly against the frontier, priced at $2/$10 per million tokens with an Artificial Analysis score of 55. (kimi.ai)

2026-10-12

  • Anthropic cuts Claude distillation attacks by 90-99%
    Anthropic announced a multi-layered defense against distillation attacks on Claude—preserved thinking, a two-stage constitutional classifier, and rapid account bans—that reduced reasoning tokens extracted by distillers by roughly 90 to 99 percent. The post also alleged that Zhipu distilled US models' cyber capabilities for GLM-5.3 and that Moonshot's Kimi secretly forwarded user requests (including one linked to China's PLA) to Claude for distillation. (anthropic.com)
    Anthropic Zhipu AI

September 2026

2026-09-30

  • OpenAI disrupts distillation campaign tied to Moonshot AI
    OpenAI disclosed that it disrupted a coordinated adversarial distillation campaign that extracted protected model reasoning through novel cross-conversation manipulation techniques, peaking at 16,000 requests from over 4,000 users on July 24-25 and spanning a cluster of more than 15,000 accounts fully disrupted by July 28. The company attributed a core cluster of the activity to individuals associated with Moonshot AI, the developer of Kimi, and shared findings with industry partners via the Frontier Model Forum and government channels, warning that distillation poses safety and national security risks. (openai.com)
    OpenAI

August 2026

2026-08-03

  • SemiAnalysis Deep Dive: Kimi K3 Architecture Explained
    SemiAnalysis published a comprehensive technical breakdown of Kimi K3's architecture, covering its Kimi Delta Attention (KDA) hybrid attention mechanism, Attention Residuals for cross-layer information flow, Stable LatentMoE with quantile load balancing, and inference optimizations for agentic workloads. The analysis reveals K3's unconventional design choices, such as a 3:1 KDA to Multi-Head Latent Attention ratio, block attention residuals for reduced communication overhead, and FlashKDA kernels that enable linear-time prefill and constant-time decode. (newsletter.semianalysis.com)

2026-08-02

  • Open Model Roundup: Laguna S2.1, Inkling, Kimi K3
    Interconnects' August 2026 roundup of open model releases highlights how model training capacity is proliferating rather than consolidating, with new releases from Thinking Machines (Inkling), Tencent (Hy3), Poolside (Laguna S2.1), DeepSeek (V4-Flash), and Moonshot (Kimi K3), alongside other entries from Meituan, AMD, and Swiss-AI. The recap underscores that more organizations are openly releasing strong models despite rising training costs, with notable trends including Tencent's shift to Apache 2.0 licensing and Kimi K3's restrictive revenue-share license raising potential U.S. policy implications. (interconnects.ai)
    Poolside Thinking Machines Lab DeepSeek

July 2026

2026-07-27

  • Moonshot AI Releases Kimi K3 Model
    Moonshot AI released Kimi K3, a 2.8 trillion parameter open-source frontier model that ranks third globally on the Artificial Analysis Intelligence Index. The model adopts an 'affordable luxury' pricing strategy that sets a high price ceiling for domestic Chinese models to shift the market away from cutthroat price competition toward value monetization. (kimi.com)

2026-07-16

  • Moonshot AI releases frontier open-weights model Kimi K3
    Moonshot AI released Kimi K3, a 2.8 trillion parameter Mixture-of-Experts model with open weights planned for July 27th, which ranks among the top frontier models globally. The release significantly narrows the performance gap between open and closed AI models, highlighting China's growing capital efficiency and commitment to open-source AI. (interconnects.ai)

February 2026

2026-02-24

  • Anthropic accuses Chinese labs of distilling Claude
    Anthropic accused Chinese AI labs DeepSeek, Moonshot, and MiniMax of conducting industrial-scale distillation campaigns using thousands of fraudulent accounts to extract knowledge from its Claude models, raising national security concerns. An in-depth analysis of the campaigns suggests the impact varies significantly, noting that while MiniMax and Moonshot generated millions of exchanges, DeepSeek's attributed activity was comparatively small. (go.theregister.com)
    Anthropic DeepSeek MiniMax

January 2026

2026-01-27

  • Moonshot AI releases Kimi K2.5 open weights model and Kimi Code
    Moonshot AI released the open-source Kimi K2.5, a native multimodal model trained on 15 trillion tokens that challenges closed models in coding and video benchmarks. The company also launched Kimi Code, a coding agent that allows developers to use images and videos as input within terminals or IDEs. (techcrunch.com)

2025

November 2025

2025-11-28

2025-11-12

  • Moonshot AI releases Kimi K2 Thinking model
    Moonshot AI introduced Kimi K2 Thinking, an enhanced open-source reasoning agent that handles step-by-step problem-solving and chains hundreds of tool calls, edging out top rivals on benchmarks. This builds on their push for agentic intelligence with strong long-context retention. (turingpost.com)

2025-11-10

  • Kimi-Linear attention architecture unveiled
    Kimi-Linear emerges as a hybrid design blending linear recurrence with selective full attention for efficient long-context modeling. It promises to overcome quadratic scaling limits while preserving expressive power through innovations like Kimi Delta Attention. (vizuara.substack.com)