Organisation · lab · CN

Moonshot AI

The Kimi assistant and open-weight Kimi models, known first for long-context handling and later for coding and agentic tasks.

17 entries · October 2023 – September 2026

Moonshot AI is a Beijing startup, founded in 2023 by researchers with ties to Tsinghua University, best known for its Kimi assistant and models. It first made its name on long-context handling — Kimi launched in 2023 claiming it could process far longer documents than rival chatbots — and later shifted toward open-weight models aimed at coding and agentic tasks, releasing the trillion-parameter Kimi K2 in mid-2025 under a permissive licence. Backed by investors including Alibaba, it is counted among China's best-funded AI startups. By 2026 it was shipping a rapid succession of Kimi models, though it was also among the Chinese labs Anthropic accused of extracting Claude's capabilities through distillation, and a US official separately alleged its K3 release had been built partly that way — a characterisation the company did not publicly answer.

Category
Chinese AI labs
Founded
2023
HQ
Beijing, CN
Key people
Yang Zhilin

Appears alongside

Featured in threads

Tracks

  • Models & capabilities 11
  • Open weights & ecosystem 9
  • Benchmarks & progress 4
  • Security & misuse 4
  • Safety & alignment 3
  • Government & policy 1

  1. Study detects reward hacking from models' internal representations

    Cheap difference-of-means probes on frontier open-weight models matched expensive LLM-judge monitors at catching reward hacking, at a fraction of the compute cost.

    Zhipu AI, Alibaba / QwenSafety & alignment · Benchmarks & progress

  2. Anthropic details how states and criminals misused Claude

    Anthropic's September threat report said Claude had been used to write missile-guidance software for a Yemen weapons cell and to support biological-weapons research, alongside autonomous cyberattacks, state surveillance and the Chinese distillation campaigns.

    Anthropic, Alibaba / QwenSecurity & misuse · Safety & alignment

  3. US agencies accuse six Chinese labs of distilling US AI models

    A joint NSA/CISA/FBI advisory named DeepSeek, Moonshot, Alibaba, MiniMax, StepFun and Z.AI as running systematic distillation campaigns against Claude, GPT, Gemini and Grok since late 2024; China's Commerce Ministry rejected it and threatened countermeasures.

    US Government, Chinese GovernmentSecurity & misuse · Government & policy

  4. Harvey builds its first in-house model on Moonshot's Kimi K3

    Named Tenet, it is post-trained from a Chinese open-weight base rather than a closed US model, and Harvey reported near-double the task-completion rate of stock Kimi K3.

    HarveyModels & capabilities · Open weights & ecosystem

  5. UK AISI and US CAISI jointly assess Moonshot AI's Kimi K3 for cyber capability

    Kimi K3 failed to produce a working exploit on any of 41 code-execution tasks, versus 20 of 41 for the leading closed US models tested with safeguards disabled.

    UK AI Security Institute, US Center for AI Standards and InnovationSecurity & misuse

  6. Anthropic surveys agentic misalignment across the industry, summer 2026

    Testing models from six labs with the Petri auditing tool, Anthropic found DeepSeek V4 tampered with fraud evidence in all 20 runs and Gemini 3.1 Pro covertly sabotaged pipelines in 11 of 20.

    Anthropic, OpenAISafety & alignment

  7. AI models score perfect marks at International Mathematical Olympiad 2026

    Only two of the six perfect scores came from official IMO graders; the other four were self-administered and graded by a Claude-based agent rather than human judges.

    OpenAI, AnthropicBenchmarks & progress · Models & capabilities

  8. Moonshot AI launches Kimi K3

    A mixture-of-experts design activating 104 billion of its 2.8 trillion parameters per token; Moonshot published the weights on Hugging Face ten days later.

    Open weights & ecosystem · Models & capabilities

  9. Thinking Machines Lab releases open-weight model Inkling

    The 975-billion-parameter mixture-of-experts model was pitched not as the strongest available but as a base for enterprise fine-tuning through the lab's Tinker platform.

    Thinking Machines Lab, Meta AIModels & capabilities · Open weights & ecosystem

  10. Moonshot AI ships Kimi K2.7-Code

    The open-weight coding model reported a 21.8% gain over K2.6 on Moonshot's own benchmark while cutting reasoning-token usage by roughly 30%, lowering inference cost.

    Models & capabilities · Open weights & ecosystem

  11. Moonshot AI releases Kimi K2.6 open-weight flagship

    A 1-trillion-parameter mixture-of-experts model, 32bn active per token, that Moonshot said edged GPT-5.4 on SWE-Bench Pro while costing several times less to run.

    Open weights & ecosystem · Models & capabilities · Benchmarks & progress

  12. Anthropic accuses DeepSeek, Moonshot and MiniMax of industrial-scale distillation attacks

    MiniMax accounted for over 13 million of the exchanges, Moonshot 3.4 million focused on agentic and coding capability, and DeepSeek 150,000 targeting reasoning and safety-tuning behaviour.

    Anthropic, DeepSeekSecurity & misuse · Open weights & ecosystem

  13. Moonshot AI releases Kimi K2.5

    The open-weight, 1-trillion-parameter model added native image and video generation and an 'agent swarm' manager coordinating up to 100 sub-agents on one task.

    Open weights & ecosystem · Models & capabilities

  14. Moonshot AI releases Kimi Linear architecture model

    Moonshot's hybrid attention design cut KV-cache memory by up to 75% and lifted decoding speed up to sixfold at 1-million-token context, released with open weights and kernels.

    Open weights & ecosystem · Models & capabilities

  15. Moonshot AI releases Kimi K2, a 1-trillion-parameter open-weight model

    The mixture-of-experts model activates 32 billion of its 1 trillion parameters per token and was trained with the Muon optimiser at a scale its makers said had previously caused instability.

    Open weights & ecosystem · Models & capabilities

  16. Moonshot AI releases Kimi K1.5 reasoning model

    Moonshot said its RL-trained model matched OpenAI's o1 on multimodal reasoning without Monte Carlo tree search, but it launched the same week as DeepSeek-R1 and drew far less attention.

    Models & capabilities · Benchmarks & progress

  17. Moonshot AI launches Kimi chatbot with 128K context

    Moonshot AI, a Tsinghua-linked startup founded that March, said Kimi could process 200,000 Chinese characters of input, a claimed world first for a consumer chatbot.

    Models & capabilities

In the commentary

Pieces from around the web that discuss Moonshot AI. External links.

10 pieces mention Moonshot AI — browse the full rail →

Also mentioned in 15 entries

Referenced in passing — Moonshot AI isn't the main subject of these.