Meta Platforms

Artificial Intelligence · Large Language Models

Launch

Meta launches Muse Spark 1.3, says it now rivals OpenAI and Anthropic

Independent benchmarks back some of the bravado, ranking the model just behind Claude's top variants and ahead of GPT-5.6.

• Meta released Muse Spark 1.3 on September 2, 2026, its fourth Muse Spark model in five months, available to developers via its API and rolling out soon to Facebook, Instagram and Meta AI.
• Meta Chief AI Officer Alexandr Wang called it the company's biggest jump in model performance yet, driven by gains in coding and agentic task automation.
• Wang said the model is competitive with Anthropic's Claude Fable 5.1, beats OpenAI's GPT-5.6 Sol at coding, and outperforms current Chinese models, while OpenAI's next model, Astra, is still to come.
• Independent evaluator Artificial Analysis scored Muse Spark 1.3 at 61-62 on its Intelligence Index, trailing only Claude Fable 5.1 and Claude Opus 5 and tying GPT-5.6 Sol and Grok 4.6.
• The model uses about 25% fewer tokens per task than its predecessor, asks clarifying questions on ambiguous prompts, and confirms before taking irreversible actions.
• Meta has not decided whether to release Muse Spark 1.3's weights, even as CEO Mark Zuckerberg promised an open-weights version of the flagship model 'soon.'
• Meta remains on track developing a larger model codenamed Watermelon, its next attempt to close the gap with frontier labs.
• The release shows Meta translating its massive AI infrastructure spending into models that independent benchmarks now place among the top tier, tightening a race where lead times between labs are shrinking to weeks.