Claude Opus 5 drops to conditional as Opus 5.5 lists at $4/$20

Verdict: Claude Opus 5 moves from recommended to conditional. Keep it where it is already validated. Do not start new workloads on it. Its successor, Claude Opus 5.5, is 20% cheaper per token, but it sits at conditional in our directory too, so test before you switch.

What happened

We rechecked Anthropic's live pages on 31 October 2026. They show three facts:

• Opus 5.5 lists at $4 input and $20 output per 1M tokens. Opus 5 lists at $5/$25 on the same pricing page, so Opus 5.5 is 20% cheaper on both sides. Cache reads drop from $0.50 to $0.20 per 1M tokens, 60% less. Anthropic's launch post also says that "at default settings it will cost 40% less than Opus 5 on typical workloads"; that figure is Anthropic's, not ours.
• Anthropic's models overview lists Opus 5 under "Legacy models (still available)" and lists Opus 5.5 in the current lineup. Legacy means you can still call it, and Anthropic's model deprecations page commits to no Opus 5 retirement before 24 July 2027.
• Anthropic's own guidance is explicit: "If you're unsure which model to use, start with Claude Opus 5.5 for most workloads."

Why it matters

When we rated Opus 5 recommended on 25 July 2026, the case was near-Fable 5 intelligence at half the Fable price. That case now belongs to a cheaper successor from the same vendor, and the vendor itself points new work away from Opus 5. A recommended rating on a legacy model that costs more than its replacement no longer holds.

We have not run Opus 5 against Opus 5.5 on our own workloads, so we make no claim about relative quality. The price gap and the legacy label are the evidence for this verdict, and both come straight from Anthropic.

Opus 5.5 is not an automatic upgrade either. We rate it conditional because every benchmark behind it is Anthropic's own and migration is not drop-in: Anthropic's overview lists its thinking as "Adaptive (always on)", and Anthropic publishes a dedicated migration guide for teams coming from Opus 5 or earlier.

What changes for you

• Running Opus 5 on validated pipelines: no forced move. Re-run your evals on Opus 5.5 and take the 20% per-token saving if they pass.
• Starting new work: do not start on Opus 5. Start your evals on Opus 5.5, Anthropic's recommended default.
• Weighing Claude Fable 5.1 at $10/$50: Anthropic's models overview points to Fable 5.1 for demanding reasoning and long-horizon agentic work, or when your evals on Opus 5.5 at higher effort still fall short.

FAQ

Is Claude Opus 5 being shut down? No. Anthropic lists it as legacy and still available, with retirement not sooner than 24 July 2027.

Is Opus 5.5 better than Opus 5? Anthropic says so, but we have not tested it ourselves. The verdict rests on price and Anthropic's legacy label, not on our benchmarks.