Claude Opus 5

The quiet change in this row is a boolean. Opus 5 ships with reasoning enabled by default — the catalog's reasoning.default_enabled reads trueopenrouter-models, last checked 2026-09-14 here, where the anthropic/claude-opus-4.7 and anthropic/claude-opus-4.8 rows both carry false. The effort ladder itself is unchanged from those releases (low, medium, high, xhigh, max, defaulting to highopenrouter-models, last checked 2026-09-14); what changed is which setting you get if you say nothing, and therefore what a naive request costs.

Speed is a second dial with a published exchange rate. The separate anthropic/claude-opus-5-fast row lists not publishedlast known value, as of an unrecorded date — the source no longer lists this rowopenrouter-models, last checked 2026-09-14 input against this row's $5.00per million tokensopenrouter-models, last checked 2026-09-14 for what Anthropic's launch post describes as around 2.5x the default speed, at twice the base pricesource, accessed 2026-08-28. Going the other way, anthropic/claude-opus-5:batch halves the price for work that can wait, and a cache hit reads at $0.50per million tokensopenrouter-models, last checked 2026-09-14. Between a cached batch read and fast-mode output, the price of a token from one model spans more than two orders of magnitude.

The comparison Anthropic drew was with its own dearer tier. Opus 5 lands "within 0.5% of Fable 5's peak score" on Cursor's coding benchmark at half the cost per task, and beats Fable 5's best OSWorld result "at just over a third of the cost." The catalog's scoreboard agrees and goes further: the Artificial Analysis intelligence index on this row is 50.7openrouter-models, last checked 2026-09-14 against 49.7openrouter-models, last checked 2026-09-14 on anthropic/claude-fable-5, which lists at $10.00per million tokensopenrouter-models, last checked 2026-09-14 input — the dearer of the two. On that measure the premium tier is behind the flagship and ahead of it on the invoice.

One capability is explicitly absent. The same post says Opus 5 "remains behind Mythos 5 on cybersecurity tasks" — the tier Anthropic does not sell on the open market, which returned from a June 2026 suspension only after US government approval. Whatever ceiling this row represents, it is the top of the purchasable stack rather than the top of the stack.