Poolside Releases Open Weights for Laguna M.1, a Coding Model With 256K Context
Poolside released the weights for Laguna M.1, an open-weight agentic coding model with a 256K context length, making base and post-trained checkpoints available on Hugging Face under Apache 2.0. Support was immediately available in vLLM v0.21.0 and SGLang.
vLLM described Laguna M.1 as a 70-layer sparse mixture-of-experts model with 225 billion total parameters, 23 billion active per token and 256 experts with top-k=16 routing. SGLang highlighted native interleaved reasoning between tool calls and results on SWE-bench Verified, SWE-bench Multilingual, SWE-Bench Pro and Terminal-Bench 2.0, while a later post described a 3-bit MLX build for running the model on a Mac.
From the sources (10 posts)
@lmsysorg🎉 Day-0 support for Laguna M.1 from @poolsideai is live in SGLang! This is a 225B MoE built for agentic coding & long-horizon work. 1️⃣ 70-layer MoE: 3 dense SwiGLU layers + 67 sparse MoE layers, 256 experts, top-k=16 with aux-loss-free loa
@poolsideaiToday we’re releasing the weights for Laguna M.1, our most capable model to date, with a 256K context length. Both base and post-trained checkpoints are now available on Hugging Face under Apache 2.0.
@_lewtunRT @poolsideai: Today we’re releasing the weights for Laguna M.1, our most capable model to date, with a 256K context length. Both base an…
@vllm_project🎉 Congrats to @poolsideai on Laguna M.1, a new open-weights agentic coding model. Day-0 support landed in vLLM v0.21.0. 🧠 70-layer sparse MoE: 225B total params, 23B active per token, 256K context 🔀 256 experts with top-k=16 routing, built
@eisokantRT @poolsideai: someone asked if they could try getting Laguna M.1 running on a Mac. we said yes. they came back with a 3-bit MLX build r…
@clementdelangueRT @poolsideai: someone asked if they could try getting Laguna M.1 running on a Mac. we said yes. they came back with a 3-bit MLX build r…
@clementdelangue"Open weights are now our default"
@eliebakouchbase model is also available, very nice!
@huggingfaceRT @poolsideai: Today we’re releasing the weights for Laguna M.1, our most capable model to date, with a 256K context length. Both base an…
@victormustarRT @poolsideai: Today we’re releasing the weights for Laguna M.1, our most capable model to date, with a 256K context length. Both base an…