Meta unveils advanced AI model muse spark 1.3 with enhanced coding capabilities

Meta Platforms announced on September 2, 2026, the release of Muse Spark 1.3, its most powerful AI model to date. Designed to extend the workflow of intelligent agents and enhance coding abilities, the model outperforms OpenAI's GPT-5.6 Sol and matches Anthropic's Claude Fable 5.1 in performance. The pricing for Muse Spark 1.3 remains the same as its predecessor, with a cost of $1.25 per million input tokens and $0.15 per million cached input tokens. Meta's Chief AI Officer, Alexandr Wang, highlighted the model's efficiency, noting a 25% reduction in token usage for similar tasks and a 20% decrease in tool calls. The model is being integrated into Instagram, Facebook, and Meta AI Assistant, with some developers already using it at a weekly rate of trillions of tokens. Muse Spark 1.3 is optimized for long-term intelligent agent tasks, enabling collaboration across multiple workflows in a single thread. It actively identifies plan gaps, tracks learning outcomes, and seeks user clarification when faced with ambiguous instructions. The model also improves reliability in handling complex, multi-step tasks by preserving detailed requirements and constraints. Enhanced multi-tasking capabilities allow it to accurately map new instructions to the correct task within a single thread, while better self-awareness reduces hallucination outputs.