Meta says its new AI model is ready to compete on coding
TL;DR
Meta is opening Muse Spark 1.1 to US developers through the new Meta Model API in public preview. New accounts get 20 dollars in free credits. The model is pitched as stronger at coding, including finding and fixing complex bugs, supporting end-to-end agentic workflows, and handling multi-agent systems. Muse Spark 1.1 is also available in Thinking mode inside the Meta AI app and website. Meta calls it a step change from the first Muse Spark model launched in April.
Nauti's Take
Meta is saying: take us seriously as a developer AI platform, not only as the company putting chatbots into social apps. That makes sense strategically, but the burden of proof is still open.
The useful test is not whether Muse Spark 1.1 sounds good in launch language, but whether it handles real repositories, messy bugs, tool calls, and agent chains better than the models developers already trust. Until benchmarks and hands-on reports land, this is promising but PR-heavy.
Briefingshow
Meta is trying to move Muse Spark from consumer AI surfaces into developer workflows, where the highest-value AI usage is happening. Coding agents live or die by reliability, tool access, latency, price, and debuggability. Meta’s claims are ambitious, but developers will judge the API against OpenAI, Anthropic, Google, and specialized coding models in daily work, not launch copy.
Sources