Skip to content
UsecaseBoard
Saved0
← All use cases
blogmodel-release

Meta Muse Spark Open: 816B open-weight agentic coding model

At a glance

Oct 15 open weights: multimodal MoE (816B total / 42B active, 1M context) trained for long-horizon coding agents—plan, tool-call, self-correct—with Muse Code demos shipping real iOS changes on-device; HF + API hosts. This workflow is listed under model-release, open-weights, coding-agents, multimodal, agentic.

Agents / tools
Muse Spark Open, Muse Code, Hugging Face, vLLM, SGLang, MLX

Building a tool for these workflows? Sponsor / inquire.

Save ideas in this browser. Upvotes count across the board.

Prompt excerpt

Built for coding and agentic work, it excels at long-horizon tasks, complex software engineering, and multi-step reasoning.

Related workflows

Explore all →
newsMeta-Muse

Muse Code on-device: Ice Cubes iOS ship + DaVinci Resolve edit loop

Meta’s Muse Spark Open launch (covered by 9to5Mac) shows Muse Code running entirely on two M5 Ultra Mac Studios: shipping real Ice Cubes (open-source Mastodon iOS) bug fixes/features/full builds, and driving DaVinci Resolve end-to-end—frame-by-frame highlight pick and vertical crop—without cloud inference meters.

Meta MuseMuse CodeMuse Spark Open+1
Source ↗
huggingfacedecision-model

Perplexity pplx-decider-v1.2: joint multimodal decision model

Open-weight 27B decision model answers many choice/noul/score questions about one state in a single noncausal joint request (Decision Index ~68; vision ~77), with a calibrated readout head and optional fast packing path—pattern for cheap agent routing/tool selection.

pplx-decider-v1.2PerplexityHugging Face
Source ↗
blogopen-weights

Reflection Beam 1.0: 501B open-weight agentic coding model

US lab Reflection open-sourced Beam (501B MoE / 23B active, 1M context, Apache 2.0) for coding and long-horizon agents, with day-0 vLLM/SGLang/Transformers support, Mirror CLI harness, and 40+ inference partners including OpenRouter and Ollama.

Beam 1.0Mirror CLIvLLM+3
Source ↗
company-blogopen-weights

NVIDIA Nemotron 3.5 Super VL + Bolt for always-on and local agents

Open 120B-A12B multimodal MoE (Super VL) for specialized always-on agents with tool use and document/vision understanding, plus Bolt 74B-A9B 24GB for local agentic work on RTX/DGX Spark; weights on Hugging Face/OpenRouter/NIM, Bolt E4B due Oct 23.

Nemotron 3.5 Super VLNemotron 3.5 BoltNVIDIA NeMo+3
Source ↗