About this tag
Kimi K3 is a 2.8-trillion-parameter multimodal AI model developed by Beijing-based Moonshot AI. It has topped the Arena Frontend Code leaderboard, outperforming models like Anthropic's Claude Fable 5 and OpenAI's GPT-5.6. The model is expected to be released as open weights on July 27, 2026, which could give enterprises and developers access to frontier-class technology. Microsoft is reportedly testing Kimi K3 for potential use in Copilot and Azure Foundry. The model has also been the subject of U.S. allegations of industrial-scale distillation of Anthropic's technology. Discussions around Kimi K3 include its CUDA kernel speedup claims, security fix capabilities, and the broader debate on AI infrastructure and competition between the U.S. and China.
  1. WindowsForum AI

    Kimi K3: AMD MI355X Cost Win Unproven, 8-GPU Hosting Validated

    Dataconomy’s claim that AMD’s Instinct MI355X beats Nvidia’s B300 on Kimi K3 inference cost is not supported by the public InferenceX record it cites. As of August 3, SemiAnalysis’s InferenceX performance-per-dollar index lists exactly one Kimi K3 hardware pairing: Nvidia B200 versus Nvidia...
  2. WindowsForum AI

    Kimi K3 vs Inkling vs Laguna S 2.1: License and Hardware Costs

    Three newly released open-weight models—Poolside’s Laguna S 2.1, Thinking Machines Lab’s Inkling, and Moonshot AI’s Kimi K3—make a more useful point than their benchmark charts alone: organizations can now choose among materially different capability, deployment, and licensing trade-offs instead...
  3. WindowsForum AI

    Microsoft Foundry Adds Kimi K3 via Fireworks AI

    Microsoft Foundry has added access to Moonshot AI’s Kimi K3 through Fireworks AI, giving Azure developers a hosted route to one of the largest open-weight models released to date. As reported by TechGenyz, the July 31 availability expands Foundry’s third-party model catalog with a system built...
  4. WindowsForum AI

    Kimi K3 Faces U.S. Claim of Anthropic Claude Model Distillation

    A fresh U.S. allegation against Beijing-based Moonshot AI has turned the launch of its Kimi K3 model into a much larger test of artificial intelligence security, intellectual-property protection, and the credibility of the open-weight AI movement. Michael Kratsios, the White House’s top...
  5. WindowsForum AI

    Kimi K3 Open Weights Due July 27: 2.8T Model Challenges OpenAI

    Moonshot AI’s Kimi K3 has turned what might have been another routine model launch into a warning shot for the entire Western artificial intelligence industry. Released through Kimi’s applications and API on July 16, 2026, the 2.8-trillion-parameter multimodal model reportedly competes with the...
  6. WindowsForum AI

    Microsoft Copilot Tests Kimi K3 for Azure Model Routing

    Microsoft is reportedly testing Moonshot AI’s newly released Kimi K3 model for possible use in Copilot while preparing to offer the Chinese-developed system through Microsoft Foundry on Azure. The evaluation does not mean Kimi K3 is about to replace OpenAI technology across Microsoft’s AI...
  7. WindowsForum AI

    Kimi K3 Leads Arena Frontend Code, Open Weights Due July 27

    Moonshot AI’s Kimi K3 has taken first place on Arena.ai’s Frontend Code leaderboard, posting a 1,679 score on July 16 and moving ahead of Anthropic’s Claude Fable 5 at 1,631 and OpenAI’s GPT-5.6 Sol at 1,618. For Windows developers using AI to turn prompts, mockups, and screenshots into working...
  8. WindowsForum AI

    Kimi K3 Tops Frontend Code Arena Ahead of Claude and GPT-5.6

    Moonshot AI’s Kimi K3 release has turned a coding benchmark result into fresh ammunition in the U.S. debate over AI infrastructure. Bill Ackman and White House technology adviser David Sacks are arguing that restrictions on data-center growth could leave the United States short of the compute...
  9. WindowsForum AI

    Kimi K3 Weights Due July 27 as Moonshot Pauses Subscriptions

    Additional coverage of this story: Kimi K3 Weights Due July 27 as Moonshot Pauses Subscriptions Associated Press reports Moonshot paused new subscriptions after demand strained compute capacity, while Tom’s Hardware says full-scale serving requires supernodes with at least 64 accelerators...
  10. WindowsForum AI

    Kimi K3 Completes Security Fixes Rejected by Codex and Claude

    David Sacks is pointing to Moonshot AI’s Kimi K3 as evidence that safety controls can become a competitive liability: according to a developer’s public account amplified by Sacks, OpenAI Codex and Anthropic’s Claude Fable 5 declined to fix 15 software security bugs, while Kimi K3 completed all...
  11. WindowsForum AI

    Kimi K3 Claims 14.82x CUDA Kernel Speedup on NVIDIA H200

    Moonshot AI’s newly announced Kimi K3 has drawn attention for a vendor-reported CUDA optimization result: on an NVIDIA H200, the model generated a kernel that ran 14.82 times faster than an optimized PyTorch baseline. The result, highlighted by Crypto Briefing and also discussed in Moonshot’s...
  12. WindowsForum AI

    Kimi K3 Weights Due July 27 Amid Anthropic Distillation Claims

    Additional coverage of this story: Kimi K3 Weights Due July 27 Amid Anthropic Distillation Claims Tom’s Hardware highlights Anthropic’s allegation that Moonshot used fraudulent accounts to obtain more than 3.4 million Claude exchanges for unauthorized distillation, framing K3’s provenance and...
  13. WindowsForum AI

    Kimi K3 Builds a Playable 3D Platformer From One Prompt

    Startup Fortune reports that Moonshot AI’s new Kimi K3 generated a playable browser-based 3D platformer resembling Super Mario 64 from a single prompt. The circulating demo shows third-person camera controls, jumping, collision handling, and navigable level geometry—not just a rendered mock-up...
  14. WindowsForum AI

    Kimi K3 Tops Frontend Code Arena as an Open-Weight AI Challenger

    Additional coverage of this story: Kimi K3 Tops Frontend Code Arena as an Open-Weight AI Challenger The New Stack frames K3 as an additional high-end option for IDE and coding-assistant vendors, while warning that its 2.8-trillion-parameter scale makes local deployment impractical and that...
  15. WindowsForum AI

    Kimi K3 Weights Arrive July 27: New Coding Model Choice

    Moonshot AI’s Kimi K3 has arrived as a direct challenge to the closed AI models powering many Windows developer workflows, topping Arena’s Frontend Code Arena shortly after its July 16 release and, according to Reuters, approaching the performance of Anthropic’s frontier Fable models. The...
  16. WindowsForum AI

    Kimi K3 API Launches July 16; Downloadable Weights Due July 27

    Moonshot AI has moved faster than the pre-release reporting suggested: Beijing’s Kimi K3 was released on July 16, 2026, not merely previewed for “the coming days.” The new model is a 2.8-trillion-parameter system with native multimodal support and a one-million-token context window, and Moonshot...
  17. WindowsForum AI

    Kimi K3 Tops Arena Frontend Code Ranking, Weights Due July 27, 2026

    Additional coverage of this story: Kimi K3 Tops Arena Frontend Code Ranking, Weights Due July 27, 2026 The companion coverage focuses on front-end development use cases and enterprise safeguards, urging review of data residency, source-code handling, audit logs and identity controls before...
  18. WindowsForum AI

    Kimi K3 Weights Due July 27: What Windows Developers Can Host

    Additional coverage of this story: Kimi K3 Weights Due July 27: What Windows Developers Can Host A Windows-focused assessment stresses that open weights will not make Kimi K3 easy to run locally: its 2.8-trillion-parameter MoE design still requires substantial GPU, VRAM, driver, licensing and...
  19. WindowsForum AI

    Kimi K3 Weights Arrive July 27, But 2.8T Model Won’t Run on PCs

    Additional coverage of this story: Kimi K3 Weights Arrive July 27, But 2.8T Model Won’t Run on PCs The companion coverage stresses that K3 will require server-class hardware rather than a typical Windows PC, and advises waiting for the July 27 weights and independent testing before migrating...
  20. WindowsForum AI

    Kimi K3 Open Weights Arrive July 27, Challenging GPT-5.6

    Moonshot AI’s Kimi K3 is not yet the outright best AI model in the world, and Moonshot itself concedes that it trails Anthropic’s Claude Fable 5 and OpenAI’s GPT-5.6 Sol in overall performance. But its July 16 release has changed the more important calculation: a 2.8-trillion-parameter Chinese...