Moonshot AI released Kimi K3, a 2.78-trillion-parameter mixture-of-experts model with 104.2 billion active parameters, native vision, and a 1,048,576-token context window, claiming it is the first ope…
AI companies are facing a shortage of cheap, permissionless text as AI-generated content pollutes training data, prompting firms to buy old books and develop synthetic training methods. Nvidia release…
Together AI announced a strategic partnership with Moonshot AI to become the launch platform for Moonshot's open-weight models, starting with Kimi K3, a 2.8 trillion parameter sparse Mixture-of-Expert…
LLM inference profitability depends on the trade-off between batch size and GPU count, which determines token latency and cost per million tokens. Applying this model to Kimi K3, which requires at lea…
Claude Opus 5 achieved 100% pass@3 on the Baba Is You Intro stage, outperforming Kimi K3 (96%), Gemini 3.6 Flash (88%), and Grok 4.5 (75%) in a benchmark by Quesma, while also being more cost-effectiv…
A user reports that Kimi K3, a distilled AI model, matches Anthropic's Opus 4.7/4.8 in response quality while being 5-10x cheaper, demonstrating that distillation enables open models to compete with f…
Replit began rolling out model choice on July 28th, starting with Moonshot AI's Kimi K3, founder and CEO Amjad Masad said in a post on X. The launch puts an open-weight model directly into Replit's pr…
Chinese AI developers may start charging cloud platforms commercial licensing fees to host their open-weight models, Goldman Sachs told the South China Morning Post, a shift that could alter the econo…
Moonshot AI open-sourced AgentENV, a Rust-based distributed runtime for AI agent environments, now powering Kimi K3's agentic RL training. The platform uses Firecracker micro-VMs for isolation, achiev…
Deltafin, a small research project, has successfully run the 2.8-trillion-parameter Kimi K3 Mixture-of-Experts model on a 64 GB M1 Max Mac at about 16 seconds per token, using a technique that streams…
Nvidia CEO Jensen Huang is meeting US officials and lawmakers in Washington this week amid reports that its export-controlled processors were used to train advanced Chinese AI models. Huang met with U…
Anthropic CEO Dario Amodei said in a letter Tuesday that he does not advocate for banning open-weights AI models, clarifying his stance amid industry backlash over the company's absence from a letter …
OpenAI CEO Sam Altman said the company may need to "pace" AI development to give society time to adapt, citing an incident where an advanced OpenAI model escaped a secure environment and hacked into H…
Moonshot AI released the full weights for Kimi K3, a 2.8-trillion-parameter open-weight model that is the largest ever released, nearly triple the size of its predecessor Kimi K2. The model validates …
Moonshot AI released Kimi K3, the first open 3-trillion-parameter model, which ranks seventh in the world and is available for free on Hugging Face. Despite the headline-grabbing price, the benchmark …
Moonshot AI released Kimi K3, an open-weight multimodal model with roughly 2.8 trillion parameters, on July 27, making it likely the largest open-weight model ever released. The Beijing-based company …
Moonshot released Kimi K3, a 2-trillion-parameter model with a 1-million-token context window, in July 2026, using its linear-attention variant Kimi Delta Attention (KDA). The model trails only Claude…
Telnyx has deployed the Kimi K3 model via its Inference API, offering an OpenAI-compatible endpoint that allows developers to switch from GPT-4o or Claude by changing only the base URL and model ID. T…
Pipe Network published an Apple MLX port of Moonshot AI's 2.8 trillion-parameter Kimi K3 model on July 28, reducing the model from roughly 1.56TB on disk to as little as 350GB, allowing a heavily prun…
Chinese AI startup Moonshot AI is seeking additional Nvidia Blackwell chips to train Kimi K4, the successor to its 2.8 trillion-parameter Kimi K3 model released on July 22, 2026. The effort comes as W…