Meta Releases Muse Code and Muse Spark 1.2 for Long-Sequence Agentic Coding
Simon Willison · rss · 2026-08-06
Meta released Muse Spark 1.2, a coding-focused model, alongside the Muse Code agent. Muse Spark 1.2 features significant improvements in code generation, complex debugging, and end-to-end developer workflows.
This update highlights the growing importance of long-sequence agentic tool calling. Meta scaled up training compute on coding tasks, expanded environment diversity, and co-trained the model using rejection sampling and subagent optimizations. The model excels at long-horizon coding tasks like whole-repository generation and large projects.
More from coding & agent
- Dev Builds AI Digital Twin 'Hal' as Second Brain to Auto-Run Meetings and Emails — danshipper · 2026-08-06
- Claude Code CLI Update: Introduces Multi-Step Agent and Refined Grep Tool — ClaudeCodeLog · 2026-08-06
- Claude Code CLI Update: Multi-Agent Coordination and Permission Fixes — ClaudeCodeLog · 2026-08-06
- Will Vibe Coding Ruin Large Projects? Dev Warns of Architecture Collapse — Dangerous-Rhubarb-24 · 2026-08-06
- You.com Launches Answer API with Built-in Citation Verification at 93% Accuracy — PolarBearby · 2026-08-06
- AI Agents Remain in Single-Player Mode: No Autonomous Spending or Inter-Agent Collaboration Yet — GregKamradt · 2026-08-06