New AI model update coming soon
Meta is preparing a new version of its Muse Spark model, with major improvements expected in coding and agentic capabilities.
Chief AI Officer Alexandr Wang described the update as part of Meta’s effort to become more competitive with leading artificial intelligence platforms and expand its ambitions in enterprise AI.
Wang clarifies Meta’s AI agent progress
Wang discussed the update after CEO Mark Zuckerberg’s comments at a company townhall raised questions about the pace of Meta’s progress in AI agents.
“..Our next Muse Spark update is coming soon. Big improvements in coding and agentic capabilities to be more competitive with other leading models,” Wang wrote in a post on X.
Watermelon uses more compute
The upcoming Muse Spark update is codenamed Watermelon.
During the townhall, Wang said the model uses far more computing power than its predecessor.
He also reportedly told employees that Watermelon has already caught up with OpenAI’s flagship GPT 5.5 model, according to a report citing anonymous sources.
Enterprise AI market comes into focus
Analysts say stronger coding and agentic tools could make Muse Spark more relevant for enterprise customers.
Pareekh Jain, principal analyst at Pareekh Consulting, said a stronger Meta model would add pressure to the market and give companies another option beyond OpenAI and Anthropic.
“A strong Meta model would increase competition, lower AI costs, and give enterprises another alternative to OpenAI and Anthropic,” Jain said.
Potential cost and control advantages
Jain said the opportunity could be especially significant if Meta makes the model available as open-weight or at a lower cost.
“If offered as an open-weight or low-cost model, it could make AI coding assistants more affordable while improving data control and reducing vendor lock-in,” Jain added.
The comments reflect growing pressure in enterprise software development, where AI coding assistants are becoming more widely used but increasingly expensive to access.
GPU shortages and model costs matter
Companies adopting AI development tools are facing higher costs linked to GPU shortages, model licensing fees and inference expenses.
Those pressures have made the most capable coding models harder to scale across large organizations.
A competitive Meta model could therefore appeal to enterprises looking for more affordable and flexible AI infrastructure.
Speculation around developer tools
The timing of the Muse Spark update has also fueled speculation that Meta may be preparing a broader AI-assisted development platform.
Recent acquisition efforts, including Meta’s reported interest in Manus, have led some observers to believe the company could move toward application-building tools or vibe coding products.
Meta may move beyond foundation models
Charlie Dai, principal analyst at Forrester, said Meta appears to be positioning itself as more than a foundation model provider.
“It seems, especially with these updates, Meta wants to move beyond foundation models and become a platform for building AI-native applications and agents,” Dai said.
He said initiatives such as Pocket, even if consumer-facing, point to Meta’s interest in lowering the barrier to creating AI-native software.
Business users could become a key audience
Dai said the larger opportunity may be in enterprise adoption.
According to him, Meta could target business users who want to build workflow automations, agents and lightweight applications without needing deep technical expertise.
That would place Meta more directly in competition with enterprise AI platforms focused on productivity, automation and software development.
Cloud infrastructure strategy expands
Meta’s broader enterprise AI push may also include new cloud infrastructure business lines.
The company is reportedly developing plans to sell access to AI computing power and models, potentially turning its own AI infrastructure into a commercial product.
Execution risks remain
Analysts cautioned that winning enterprise customers will require more than a powerful model.
“Meta must prove superior real-world coding quality, reliable agent execution, strong security and governance, and a vibrant developer ecosystem,” Dai said.
He added that outside North America, geopolitical and regulatory factors are increasingly shaping model choices and creating room for alternatives.
Rollout expected through Meta AI and API
Dai said Meta will need strong customer results, local partnerships and sustained innovation to appeal to developers and enterprises.
Wang said the new model will be released soon through Meta AI and a new API.
