Semianalysis reports Mythos 2 is done training and Anthropic are not releasing the model.
Anthropic’s August 2026 Risk Report describes an unreleased internal model it calls Model 2. It is more capable than Mythos 5, with no current plan to release it externally and the normal external-release evaluation suite not completed. They describe Model 2 as a noticeable improvement on Mythos 5 for internal work. The gain falls short of the earlier leap from Opus 4.6 to Mythos Preview.
Other sources say Mythos 3 is currently a reported internal trajectory, not an announced product.
The likely Anthropic development pattern is likely –
– deploy Model 2 internally under stronger controls
– use it for coding, data generation, agent workflows, and evaluation
– collect evidence from internal use and selected external partners
– improve the next training and post-training cycle
– decide whether the successor can be released broadly, gated, or split into multiple product variants.
The critical question is whether safety evaluation and governance can improve as quickly as the models assisting the development process.
Mythos was first exposed through a CMS misconfiguration on March 26, 2026, then formally announced as Mythos Preview on April 7–8 — the first time in nearly seven years a leading lab publicly withheld a model over safety concerns. Access runs through Project Glasswing, invitation-only for 12 founding organizations and roughly 40 vetted critical-infrastructure operators, with a 244-page system card.
Unreleased OpenAI Models
Astra is OpenAI next major model. It remains unreleased as of today. On August 1, 2026, an internal version produced new results on ten long-standing open problems in mathematics and theoretical computer science, with Lean certificates published. Six days later OpenAI said it had suspended work on some aspects of Astra after an internal review found it reached the critical cybersecurity threshold”— meaning it could independently identify and carry out attacks against well-protected real-world systems.
AI spend at SemiAnalysis runs around $10M-ish annualized and has been roughly flat through Q2 after spiking in Q1. Everyone got cloud code psychosis and then it leveled out. AI spend is one-time R&D bursts, not steady state. Their repo count went from 10 to 150+. Fable’s launch didn’t move spend up. An intern was burning ~$8K/day for four days and turned out to be as productive as full-timers. They’re considering AI-run performance reviews.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
My contention is that if AI requires a safety analysis then it’s a failure, that is with something as capable as agentic AI it should be inherently safe or it not at all ever safe, period.