Open Weight
Topic archive • 9 matches
AI-generated: summaries written by AI from the linked sources. How we use AI
2026-10-15
Technology
StepFun released the open weights for Step 5 Preview, a 600B total parameter model with 27B active and a 1M-token context, along with a technical blog detailing its 92-layer sparse GQA architecture and long-horizon agent training.
AI Models • Pandaily
PermalinkAnt Group's inclusionAI open-sourced Ling-3.1-flash weights on Hugging Face under the MIT license. The 560B total parameter hybrid-linear MoE model has 25B active parameters per token and a 1M-token context window.
AI Models • Pandaily
PermalinkMeta and Apple have partnered to launch Muse Spark Open, an open-weight AI model that runs locally on Apple silicon Macs.
AI Models • TechCrunch AI
Permalink
2026-10-16
Technology
China plans to restrict exports of 23 key AI technologies, including model weights, source code, and training methods for systems hitting 10^26 FLOPs or 2 trillion parameters. A fast-track rule will still clear open-source releases as soon as they are registered.
Policy • The Decoder
PermalinkMeta announced the Muse Spark Open AI model, claiming it's the 'most capable US open-weight model to date' and the first in its class that can run on desktop using two Apple M5 Ultra Mac Studios with the MLX framework. The collaboration marks a rare partnership between Meta and Apple.
AI Models • The Verge AI
PermalinkBlack Forest Labs released FLUX 3 Image Klein, a 5.6-billion-parameter open-weights image model that supports native 4K generation, pixel-precise editing, and generation with up to ten reference images. The model runs on consumer GPUs with just 8 GB of VRAM.
AI Models • The Decoder
Permalink
2026-10-17
Technology
China plans to tighten export controls on 23 key AI technologies, including model weights, source code, and training methods for systems hitting 10^26 FLOPs or 2 trillion parameters. A fast-track rule will still clear open-source releases as soon as they are registered.
Policy • The Decoder
PermalinkDeepSeek released DeepSeek v4.1 Flash 1017, a new open-weight model available on OpenRouter and Hugging Face. Hacker News commentators noted this model is why OpenAI and Anthropic want regulatory hurdles, emphasizing the need for diversity in the AI ecosystem.
Models • Hacker News
Permalink