Researchers from UC Berkeley and UT Austin released FreeToken, an edge-native Mixture-of-Experts serving engine that runs the 753B-parameter GLM-5.2 model on a single workstation GPU, achieving 35B at…
Researchers introduced VideoGAIA, a benchmark for agentic video understanding that requires multimodal large language models (MLLMs) to perform multi-turn, tool-augmented interactions. The benchmark c…
DeepSeek released the DeepSeek-V4-Pro-0813 model weights on August 13, 2026, and Fixstars deployed it to a single-node NVIDIA B300 x8 machine within 24 hours, achieving 17,478 tokens per second throug…
Alibaba's Qwen team released the open weights for Qwen3.8-2.4T-A95B, a 2.4T-parameter MoE flagship, on August 13, 2026, and Fixstars deployed it on a single 8x NVIDIA B300 node, achieving twice the th…
A new paper from Hugging Face, 'Efficient Knowledge Distillation for LLMs: Offline Top-K Logits and a Fused Chunked KL Loss,' cuts the VRAM required for knowledge distillation from roughly 250GB to ne…
A developer experimenting with local-LLM agentic AI workflows found that combining tool calls—such as having a single `read_blog_post` MCP tool internally perform `find_blog_date`—reduces the number o…
Fireworks AI, a Series D company valued at $17.5 billion, is hiring an AI Product Engineer for its Fireworks Nexus platform, which routes AI workloads from expensive proprietary models to open models …
DeepSeek's V4-Flash-High open-weight model scored 1586 on arena.ai's Frontend Code Arena Pareto Frontier, ranking #7 overall and #3 among open-weight models, a 154-point jump from the previous V4-Pro-…
Moonshot AI released the 2.8-trillion-parameter Kimi-K3 open-weight model on July 27, 2026, and Fixstars successfully ran inference on a single-node NVIDIA B300 x8 system using SGLang's DCP support, a…
Moonshot released Kimi-K3, a 3-trillion-parameter mixture-of-experts model, on Hugging Face under an open-weights license with a revenue-sharing clause above $20 million in annual revenue, and Firewor…
Moonshotai released Kimi-K3, a 2.8 trillion parameter open-weight multimodal agentic model on Huggingface, claiming it is the world's first open 3T-class model. It features a 1-million-token context w…
Moonshot AI released Kimi-K3 with a new license that requires large Model-as-a-Service businesses (those exceeding $20 million in aggregate revenue over any consecutive 12 months) to enter a separate …
Moonshot AI has released Kimi-K3, an image-text-to-text model under a permissive license that allows use, modification, and commercial distribution, with support for Transformers, vLLM, SGLang, and Do…
Major U.S. AI companies and CEOs issued a group statement supporting open-weight models after China's Kimi-K3 open-weight model alarmed the U.S. AI industry, in what appears to be an effort to prevent…
A Reddit user identifying as ComputeIQ announced on July 21st that they are dropping Anthropic's Claude for Kimi's K3 model, citing Claude's rising per-token cost, opaque usage limits on the Fable-5 t…
China's Moonshot AI announced its Kimi-K3 open-weight model, which outperformed Anthropic's Fable 5 and OpenAI's GPT-5.6 Sol on the Frontend Code Arena benchmark, intensifying the global AI arms race.…
China's Moonshot AI has leapfrogged all US models, including Anthropic's Fable 5, in the Frontend Code Arena blind benchmark test overseen by Arena.ai, according to a report. The achievement of its op…
Moonshot AI's Kimi-K3, a 2.8-trillion-parameter open-weight model, took first place on the human-judged Frontend Code Arena on July 15, 2026, with 1,679 points, surpassing Claude Fable 5 and GPT-5.6 S…
A Chinese AI model, Kimi-K3 developed by Beijing-based Moonshot AI, jumped from number 17 to number one on the Arena.ai Frontend Code Arena benchmark, surpassing U.S. models like Claude Fable 5 and GP…