Meta has taken another major step in the artificial intelligence race with the unveiling of Muse Spark 1.1, a new agentic and coding-focused AI model, alongside the announcement of its in-house AI accelerator chip, codenamed Iris, which is expected to launch in September.
The dual announcement signals Meta’s growing ambition to compete directly with leading AI firms such as OpenAI, Google, and Anthropic, while also reducing its dependence on external hardware suppliers.
Muse Spark 1.1 represents Meta’s latest effort to push AI systems beyond simple conversational capabilities toward autonomous, task-oriented intelligence.
The model is designed with strong coding abilities and agentic functionality, enabling it to perform multi-step tasks, reason through complex workflows, and execute actions with minimal human intervention. This marks an important evolution in AI development as the industry increasingly shifts toward intelligent agents capable of handling real-world operations, software engineering tasks, research activities, and digital assistance.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
One of the key highlights of Muse Spark 1.1 is its enhanced coding performance. The model reportedly demonstrates improved code generation, debugging, and software optimization capabilities, allowing developers to use it as a virtual engineering assistant.
As enterprises continue integrating AI into their software development pipelines, advanced coding models are becoming critical infrastructure for accelerating productivity and reducing development costs. The model’s agentic framework also positions it at the center of the next wave of AI innovation.
Agentic systems differ from traditional chatbots by possessing the ability to plan, make decisions, interact with external tools, and adapt to changing objectives. This capability opens the door for applications ranging from autonomous business assistants and customer service agents to research automation and intelligent workflow management systems.
Alongside the software announcement, Meta introduced its custom AI chip, Iris, which is scheduled for release in September.
The move reflects a broader industry trend in which major technology firms are increasingly designing proprietary hardware to support their growing AI ambitions. Companies such as Google with its Tensor Processing Units and Amazon with its Trainium chips have already demonstrated the strategic advantages of owning specialized AI infrastructure.
Meta’s Iris chip is expected to play a significant role in training and deploying future generations of AI models. By developing its own silicon, Meta can potentially reduce operational costs, optimize performance for its specific AI workloads, and lessen reliance on external suppliers like NVIDIA, whose GPUs currently dominate the AI hardware market.
As demand for AI computing power continues to surge, access to dedicated infrastructure has become a key competitive advantage. The announcement also highlights the intensifying competition in the AI industry.
Technology companies are increasingly realizing that success in artificial intelligence requires both advanced models and vertically integrated infrastructure. Software innovation alone is no longer sufficient; firms must also secure computing resources, data pipelines, and specialized chips capable of supporting increasingly sophisticated AI systems.
Meta’s strategy appears to be aimed at building a comprehensive AI ecosystem that combines powerful models with proprietary hardware. If Muse Spark 1.1 delivers strong performance and Iris proves capable of handling large-scale AI workloads efficiently, Meta could significantly strengthen its position in the rapidly evolving AI landscape.
As September approaches, industry observers will closely monitor how Meta’s new model and chip perform against competing offerings. Their success could reshape competitive dynamics in artificial intelligence and further accelerate the global race toward more capable autonomous AI systems and AI-native computing infrastructure.