Anthropic’s next round of AI models has surfaced through a series of leaks, revealing three names that point to a busy stretch ahead for the company: Opus 4.8, Sonnet 4.8 and Mythos 1. The leaked details suggest Opus 4.8 is built to improve efficiency and scalability, while Sonnet 4.8 adds upgrades in vision processing and coding. Mythos 1 is described as a model aimed at enterprise-level problems with advanced reasoning.
The timing matters because Anthropic is operating under tight compute constraints while facing intense pressure from OpenAI and Google. In a separate move, the company has entered into a partnership with SpaceX to secure long-term compute resources, a sign that access to infrastructure is now central to its product plans. OpenAI is expected to release GPT 5.6, Google has confirmed Gemini 3.5 Pro, and industry analysts say 4–6 major AI models could arrive globally in June 2026.
That leaves Anthropic in a race that is no longer only about model quality. The leaked roadmap shows a company trying to stretch limited resources while matching rivals that appear to have more stable and scalable compute behind them. The result is a coming stretch in which product announcements will also be a test of who can actually deliver at scale.
If the leaks hold, Opus 4.8 will matter less as a single release than as a signal that Anthropic is still pushing to stay in the fight despite the resource gap. The next question is whether partnerships and efficiency gains will be enough to keep pace when the market’s biggest model launches land almost at once.
