Anthropic looks close to shipping Claude Opus 5, and the signals now point past internal testing into limited partner hands. The sharpest artifact so far is a research entry that surfaced briefly in a coding tool's model picker earlier this month before being pulled within hours. It was listed as an early access preview with per-turn controls, safety fallbacks, a one million token context window, and an extra-high-effort setting. A matching model string was then reported on Google's Vertex quotas catalog, echoing the sequence that preceded Sonnet 5. No pricing, system card, or public model identifier exists yet, so the spec list stays provisional.
Should it land, the model would appear in the Claude apps model selector for paid tiers, in Claude Code, on the Claude Platform, and across the three cloud partners, most likely replacing Opus 4.8 rather than running beside it. The fallback detail is the telling one: routing flagged prompts down to 4.8 is the same arrangement used above the Opus tier, which hints that this release is being positioned nearer the frontier than its number suggests.
That positioning matters because Opus has been squeezed from both sides. Sonnet 5 arrived at the end of June performing close to 4.8 at a fraction of the cost, while Fable and Mythos sit above it. Subscription-inclusive Fable 5 access ended on 19 July with usage credits taking over. A stronger Opus would give Max, Team, and Enterprise customers somewhere to move without paying Mythos rates, which is the practical reason timing chatter has settled on this week.
Cadence supports the guess without proving it:
- 70 days from 4.6 to 4.7
- 42 from 4.7 to 4.8
- 56 so far since
Expectations run toward a clear step up on coding work, though Sonnet 5's reception is a reminder that they are not always met. Anthropic has said nothing.