open-weight models
Vanaxity insights tagged open-weight models - notes on AI content, SEO, GEO and AEO from Van Data Team.
Muse Spark Open for Marketers: License, Cost and Uses
Meta released the weights for Muse Spark Open on October 15, 2026, the first open version of its Muse Spark line. It has 816 billion parameters, 42 billion active per token, a 1 million token context window and native support for text, images, video and audio. Meta calls it the most capable US open-weight model and says it can run on two Apple M5 Ultra Mac Studios. In Meta's own tests it leads on agent and tool-use benchmarks but trails Chinese open models on overall scores and coding. The license is free for most uses but requires a Built with Muse Spark Open label, and hosted access starts at $1.25 and $4.25 per million input and output tokens.
Fable 5.1 vs Gemini 3.7 Flash vs Kimi K3 vs Qwen 3.8 Max
Fable 5.1 is the most capable and priciest, Gemini 3.7 Flash is the cheap fast workhorse, Kimi K3 is the strongest independently verified open-weight model, and Qwen 3.8 Max is a capable but unproven enterprise API; for marketing and content teams the right pick turns on cost, openness, and how much of each model's claim is actually verified.