Meta unveils Muse Spark 1.1 with million‑token context window and lower pricing

Meta announced the release of Muse Spark 1.1, a multimodal, agentic AI model that supports up to one‑million tokens of context, enabling prolonged tasks such as complex coding, tool use and computer‑operation across applications. The company highlighted improvements in coding performance, multimodal understanding and safety, noting a reduced hallucination rate and stronger resistance to jailbreak attempts.

Pricing for the model’s API has been set at roughly $0.26 per task, undercutting competitors such as Anthropic, OpenAI and GLM‑5.2. Meta’s CEO Mark Zuckerberg promoted the launch on X, framing it as a challenge to rivals and sparking a price war with Elon Musk’s xAI. Benchmark data from Artificial Analysis show Muse Spark 1.1 scoring 51 on the Intelligence Index and leading the coding index with a 71.3 score, marginally ahead of GLM‑5.2 and below only GPT‑5.6 Luna.

The model is currently accessible via the Meta Model API preview and integrated into the Meta AI app’s “Thinking” mode. Meta positions Muse Spark 1.1 as a step toward its “personal superintelligence” vision, offering enterprises an affordable, high‑capacity AI capable of autonomous planning, sub‑agent delegation and large‑scale code migration.