Maxwell-V1-Flash: a 35B MoE with block-sparse attention for 256K contexts
A 35B-total / 3B-active mixture-of-experts model with 256K context, built around block-sparse attention its authors claim runs 6.7x faster than dense attention at 256K.
Why it matters
A new open MoE that pairs fine-grained routed experts with a learned block-sparse attention indexer, aiming at strong reasoning and long-context work at a practical inference cost.
What you could build with it
An indie developer could self-host the 35B MoE as the reasoning backend for a long-document analysis tool, using the 256K context to hold an entire contract or research paper in one session and returning cited clause-by-clause summaries.
Does it hold up?
Too early to judge: weights appeared on Hugging Face today with 60 likes and no independent benchmarks or usage reports yet.
Built with Maxwell-V1-Flash
- Maxwell-V1-Flash surfaces in the daily AI agents digestagentic-ai-news · Daily digest flags the new 35B/3B MoE with 256K context and block-sparse attention alongside the day's other model releases.
- Maxwell-V1.0-Flash: a next-generation 35B MoE with block sparse attentionvuink · Model card summary: 35B total / 3B active parameters, 256K context, and Maxwell Block Sparse Attention bounding main attention at 4,096 tokens per KV group.
- Maxwell-V1.0-Flash: an open AI model with 74 GB of weightscert24 · News writeup confirming the Maxwell team's open 35B/3B MoE release on Hugging Face with 256K context.
Learn more
First spotted on hf: source.
More AI models
VoiceStudio: open-source, fully-local ElevenLabs alternative for voice cloning, dubbing, and audiobooksdebpalash/VoiceStudio is a desktop app that runs voice cloning, voice design, video dubbing, dictation, transcription…models · JEV 0.75Edge0-145B: open 145B MoE that runs on a single consumer GPUEdge0 open-sourced 145B and 120B latent-MoE models derived from Qwen3.8-Flash-Next, plus an inference framework that…models · JEV 0.74Strata: run a 125B model on a gaming PCFree, MIT-licensed local inference engine that runs Qwen3.8-Flash-Next (125B MoE, about 6B active per token) on…models · JEV 0.73Google AI Edge Gallery: offline speech recognition and synthesis on-deviceGoogle added fully offline speech recognition and text-to-speech to AI Edge Gallery, with ML Kit APIs on Android and…models · JEV 0.73Soup: fine-tune an 8B LLM on a 4 GB laptop GPU from one YAML fileAn open-source CLI that fine-tunes LLMs from a single YAML file, with layer streaming that trains an 8B model on a 4…models · JEV 0.73Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.