Meta marked Thursday with major updates on its future roadmap. What broke the internet was the return of Meta chief Mark Zuckerberg to X after a break of four years. The Meta chief said that the company will be releasing Muse Spark 1.1 — an agentic and coding model that will be capable of delegating tasks to parallely-running sub-agents on desktop, mobile, as well as browsers.
The other big update was reported by Reuters that claimed that Meta is preparing to start manufacturing its own AI chip dubbed “Iris” this September onwards.
The developments have made it to the headlines when Meta is knee-deep in controversies in India, touted among the biggest market for its WhatsApp and Instagram platforms. While WhatsApp’s upcoming usernames feature has sparked a privacy row in the Asian nation, Instagram has been asked to explain how secually explicit content including minors is making it to Instagram as paid advertisements.
Zuckerberg on Muse Spark 1.1
Zuckerberg intriduced Muse Spark 1.1 as an upgraded multimodal reasoning model to upgrade advanced agentic performance and handle complex coding workloads.
The AI model comes with a one million token context window which will essentially allow it to remember and process large amounts of information throughout long work sessions.
Muse Spark 1.1 has been launched as a successor to the original Muse Spark model which was released in April this year by the Meta Superintelligence Labs. It sits above Meta’s older Llama family.
Zuckerberg said Muse Spark 1.1 will be made available through its Meta Model API. This implies that developers will now be able to rent this AI service to customize their own tools and apps. For Meta, it will bring the fees to the company from developers tapping Muse Spark 1.1.
As the company plans major pivot to AI, it could use all the extra revenue it can fetch to cover the costs of building and running the data centres needed to support the compute for its superintelligence efforts. In the first quarter of 2026, Meta reported a total revenue of $56.31 billion — clocking a year-over-year increase of 33 percent.
“Muse Spark 1.1 is strongest at agentic performance, tool use, and computer use,” Zuckerberg claimed.
The move rolls Meta into a direct competition with frontier AI services being developed by AI-native companies like OpenAI and Anthropic.
Meta is now trying to position itself in the enterprise market, offering a paid developer API to compete with the GPT and Claude AI models.
While Zuckerberg has claimed that Spark Muse 1.1 would be available at a “very low price”, it has not exactly given a number as yet.
“The Meta Model API allows developers to build using Muse Spark for the first time. Our focus is on delivering strong agentic and multimodal models at very low cost. More to come soon,” he added.
Iris in the making
With Iris, Meta is preparing to join OpenAI and Anthropic in a major pivot toward custom, in-house AI chips. The aim is to break free from the stangles of market dominators like Nvidia.
The bug-testing for Iris was reportedly completed in six weeks and no major issues came to light.
Meta is partnering Broadcom to shape up the chip’s design and teaming up with TSMC for the physical fabrication.
This push for independent hardware could slash Meta’s AI computing costs — a lot of which is presently being used to procure GPUs from companies like Nvidia and AMD.
Meta also now joining the league, it appears that AI companies foraying into the $13.5 trillion chip market is catching up as an emerging trend.
Earlier in June, OpenAI launched its first “intelligence processor” called “Jalapeño” — which has also being developed in partnership with Broadcom. OpenAI said the architecture of the chip is built to optimize data flow and resource balance for near-peak performance.
Anthropic also is tapping electronics players like Samsung and Microsoft to cut chip partnership deals and reduce heavy reliance on Nvidia.
While these developments push Meta under a development-oriented limelight, the social networking giant is yet to address its operational grievances with the Indian authorities. The company has until next week to submit official explanations on India’s inquiries around WhatsApp usernames and Instagram’s minor-safety provisions.
Meta expects to spend as much as $145 billion on AI infrastructure this year alone.
