Meta enters the crowded AI coding battle with Muse Spark 1.1

Meta has stepped further into the AI coding race with the public release of Muse Spark 1.1, a multimodal model built for agentic software work. The launch puts the company in more direct competition with coding-focused AI offerings already available from OpenAI and Anthropic.

The updated model is positioned as a tool for handling complex technical tasks that go beyond simple code generation. According to Meta, Muse Spark 1.1 is capable of multi-step reasoning, coordinating complicated processes, managing digital workflows, and helping deploy new functionality inside enterprise environments.

Meta is clearly targeting business users that want AI systems to take on larger operational workloads. The company says the model is suited for jobs such as bug fixing, large-scale code migration, and other forms of automation that enterprises increasingly expect from AI platforms.

In Meta’s description, Muse Spark 1.1 performs especially well on personal agentic tasks that involve planning and orchestrating actions across multiple external apps and services.

Meta’s pitch centers on agentic coding

Rather than presenting Spark 1.1 as only a chatbot for developers, Meta is framing it as a more autonomous system that can work through layered tasks and interact with tools. That focus on agentic behavior is becoming a major battleground in AI, especially as companies look for models that can not only suggest code but also execute broader technical workflows.

The company’s messaging suggests it sees an opportunity in organizations that need help maintaining large software systems, modernizing older codebases, and reducing the manual burden of repetitive engineering work.

Pricing may be one of Meta’s strongest weapons

Meta is not the first company to push into this segment. Rivals including Anthropic and OpenAI have already spent considerable time building out similar products. Even so, Meta’s arrival matters, especially in a market where cost remains one of the most important ways to compete.

According to Reuters, Meta plans to price Muse Spark 1.1 at $1.25 per million input tokens and $4.25 per million output tokens. That places the model roughly in the same pricing range as competing systems, though slightly above Anthropic’s Claude Haiku 4.5 and OpenAI’s GPT-5.6 Luna.

For enterprise buyers evaluating multiple AI vendors, that pricing strategy could help Meta gain attention quickly, particularly if the model’s capabilities hold up in real-world development and workflow scenarios.

Zuckerberg hints that this is only the beginning

The release appears to be significant enough internally that Meta CEO Mark Zuckerberg returned to X to comment on it — his first post there in three years. His previous post dated back to July 2023, around the period when Twitter completed its transition to the X brand.

In that post, Zuckerberg described Spark as a strong coding and agentic model offered at a very low price. He also highlighted what he said were its best areas: agentic performance, tool use, and computer use.

Just as notable was his suggestion that more releases are on the way. Zuckerberg said there was “more to come soon,” signaling that Meta likely intends to expand its model lineup further.

A crowded week in AI

The Spark 1.1 announcement landed during an especially active stretch for the AI industry. Earlier in the week, Meta also introduced a new image-generation model called Muse Image. Elsewhere, SpaceXAI released a new version of Grok, while OpenAI unveiled its GPT-5.6 model family on Thursday as well.

All of that underscores how intense the competitive pace has become. AI companies are no longer just racing on model quality alone; they are also battling on pricing, tool integration, and their ability to solve practical business problems.

With Muse Spark 1.1, Meta is making it clear that it wants a meaningful role in the fast-growing market for AI coding and agentic enterprise software tools. Whether it can stand out against more established rivals will likely depend on how well the model performs at scale — and how aggressively Meta continues to follow this launch with additional releases.