Meta disclosed plans on July 1 to launch an external cloud computing service - internally called Meta Compute - offering developers two ways to access its AI infrastructure (SiliconANGLE). The first tier would offer hosted model API access to Meta’s Muse Spark model line in an arrangement comparable to AWS Bedrock, while the second would sell raw GPU capacity directly, similar to neocloud providers such as CoreWeave (Eastern Herald). The service would be backed by Hyperion, a 2,250-acre hyperscale campus in Louisiana projected to consume 5 gigawatts of power and house millions of GPUs across 11 buildings (Eastern Herald). The move follows a pattern of AI labs monetizing spare compute capacity - xAI has pursued a comparable strategy (Gizmodo). Meta’s Muse Spark, the flagship model that would anchor the API tier, is a natively multimodal reasoning model developed by Meta Superintelligence Labs with support for tool-use and visual chain of thought (Meta AI). Meta shares rose approximately 9% on the report (SiliconANGLE).
Google released Gemini 3.1 Flash-Lite Image on June 30, a lightweight image generation model targeting enterprise developers that the company internally code-named Nano Banana 2 Lite (VentureBeat). The model generates images in approximately 4 seconds at a flat rate of $0.034 per 1,000 images and is positioned as the fastest and most cost-effective option in Google’s creative model family (VentureBeat). The release accompanied the deprecation of the gemini-3.1-flash-image-preview and gemini-3-pro-image-preview endpoints on June 25, consolidating Google’s developer-facing image generation offering (Google DeepMind). Gemini 3.1 Flash-Lite Image is available immediately via Google AI Studio, the Gemini API, and the Gemini Enterprise Agent Platform (Google DeepMind).
Anthropic completed the global redeployment of Claude Fable 5 on July 1, pairing the restore with a new cybersecurity classifier trained specifically to block the jailbreak bypass technique that triggered U.S. export controls in mid-June (Marktechpost). The classifier was developed in coordination with the U.S. government and Amazon and blocks the reported exploit in more than 99% of attempts according to the company (Anthropic). Pro, Max, Team, and select Enterprise users receive Fable 5 access under 50% of standard usage limits through July 7, after which the model shifts to usage-based credits (Marktechpost). Export controls on Mythos 5 were also lifted, with access being restored on the same timeline (Anthropic).