Overview
- A concentrated wave of model launches between July 7 and July 11 saw OpenAI roll out GPT‑5.6, SpaceXAI release Grok 4.5, Meta introduce the paid Muse family, and Anthropic extend access to Claude Fable 5, compressing product cycles and focus.
- Vendors led with token‑efficiency and aggressive pricing claims: OpenAI said Sol is 54% more token‑efficient on agentic coding, SpaceXAI priced Grok 4.5 at $2 per million input and $6 per million output with a 500,000‑token context window, and Anthropic listed Fable 5 at roughly $10 per million input and $50 per million output.
- Enterprises are moving from a price‑per‑token mindset to price‑per‑finished‑task, prompting them to build model portfolios and routing rules that send high‑volume work to low‑cost models and complex single‑pass jobs to premium models.
- Privacy and regulatory pressure have risen after Meta acknowledged Muse can generate images from a public Instagram username and media reported on 'Super Sensing' glasses that could record continuously, while U.S. government reviews and voluntary Commerce Department testing have factored into staged rollouts.
- Big compute deals and large capital commitments are letting some labs subsidize low prices for now, Apple agreed to a $250 million settlement over delayed Siri AI features, and these financial pressures will shape how long deep price cuts are sustainable.