Meta's latest AI model API release has been delayed again, postponed for nearly two months with no launch date yet.

The developer interface for Meta Platforms’ latest AI model has yet to be launched, reflecting the real challenge the tech giant faces in turning its massive AI investments into commercial gains.

On June 3rd, according to The Wall Street Journal, sources revealed that Meta has repeatedly delayed the release of the API (Application Programming Interface) for its latest AI model, Muse Spark, and as of Tuesday, there was still no specific launch date.

This delay comes nearly two months after Meta’s Chief AI Officer, Alexandr Wang, publicly promised that the API was "launching soon". In response to The Wall Street Journal’s inquiry, a Meta spokesperson stated, the company is testing the API with partners and plans to release it within this month.

The delay has directly triggered doubts from outside about whether Meta can effectively monetize its huge investment in AI infrastructure. Analysts say the delay means Meta continues to miss a crucial window to compete with rivals like OpenAI and Anthropic for developer ecosystems.

Meta’s capital expenditure plan for this year is as high as $145 billion, mainly for AI infrastructure. The API is the key channel for opening its model capabilities to external developers and achieving commercialization.

Twice delayed, launch schedule remains unresolved

Sources revealed that Meta initially planned to release the companion API alongside the Muse Spark model in April. Two days after the model went online, Alexandr Wang posted on X stating the API was "launching soon" and that the enthusiasm from developers wanting to use Muse Spark in their intelligent agent framework had greatly encouraged the team.

However, the API has not been released as scheduled. Sources say, the first delay happened between April and May due to bugs discovered during testing and the need to build more supporting infrastructure. Afterwards, the release was postponed again to June. As of Tuesday, no official launch date had been set internally at Meta.

For AI companies, API release timing is critical. Usually, companies launch APIs alongside new models or within a few weeks to maximize their impact among developers. For "closed source models" like Muse Spark that cannot be downloaded externally, the API is the only way for developers to access the model’s capabilities.

Reports state that behind the API delay is Meta’s urgent need to prove its AI investment can yield sustainable commercial returns. OpenAI and Anthropic both generate revenue by selling API access to enterprise clients, who can then embed AI capabilities into their own products and tools without building models from scratch.

Meta is targeting this market as well. CEO Mark Zuckerberg has said that businesses approach Meta every week, hoping it will build API services. He also noted that monetizing through the cloud computing business is "absolutely under consideration" to utilize the company’s excess computing power.

Meanwhile, Meta last week announced new subscription services for Instagram, WhatsApp, and Facebook, and said it would begin testing paid subscription models for its Meta AI chatbot. These moves are seen as part of Meta’s efforts to find returns for its high AI spending.

Lessons from history: The Behemoth model ended without a result

This is not the first time Meta has faced setbacks in AI model releases. According to prior reports from The Wall Street Journal, Meta last year delayed releasing an AI model called Behemoth due to engineers’ failure to greatly improve its capabilities, and the model ultimately was never officially launched.

Meta then began a large-scale talent recruitment and reorganized its AI teams, appointing Alexandr Wang to lead the newly established Meta Superintelligence Labs (MSL).

A mysterious MSL division called TBD Lab subsequently developed Muse Spark. Muse Spark marks Meta’s first ever AI model released as a closed source, with no public model weights or software files, signifying a clear shift from its previously open-source Llama series models.

The model currently provides underlying support for the Meta AI chatbot and related AI features. According to internal Meta benchmarks, Muse Spark is competitive with models from OpenAI and Anthropic in most evaluation projects and is significantly ahead of xAI’s Grok.

However, since the API has not been opened, the vast majority of developers are unable to independently verify these results. Only a few specially authorized third-party evaluators had tested the model and provided scores before its release.

With yet another API delay, doubts are again rising about Meta’s ability to bridge the gap between cutting-edge AI model development and commercial implementation.

Risk disclaimerThe market contains risks; investment requires caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial status, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article are appropriate for their particular situation. Investing based on this article is at your own risk.