Anthropic’s Claude Opus 5.5 post (September 22, 2026) frames the release as the first model in the Claude 5.5 family — and as a cost cut relative to Opus 5 without stepping off the Fable-class capability band Anthropic claims for most work.
What Anthropic published for Opus 5.5
Per Anthropic, Opus 5.5 “performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.” The company positions it as a step up from Opus 5 on agentic coding, computer use, and knowledge work, while noting that at these capability levels benchmark margins are a less reliable guide to real-world differences than efficiency.
On Anthropic’s own vendor table, Opus 5.5 is listed at 66.4% on Terminal-Bench 4.0 (xhigh effort footnote). Attribute that score to Anthropic’s reported table only — this desk did not re-run the harness. Availability is stated as live on all platforms including Amazon Web Services, Google Cloud, and Microsoft Azure, with Claude Platform callers using claude-opus-5-5.
“We’re introducing Claude Opus 5.5, the first model in our new Claude 5.5 family. It performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.”Anthropic — Introducing Claude Opus 5.5, September 22, 2026
What it costs (from Anthropic’s pricing table)
These are USD rates from Anthropic’s Opus 5.5 pricing table. Do not quote them as CAD. CAD is on the invoice. Anthropic also states that at default settings the model will cost about 40% less than Opus 5 on typical workloads — a combined effect of lower list rates and fewer tokens per task in their tests.
| Line | Input / 1M | Output / 1M | Notes |
|---|---|---|---|
| claude-opus-5-5 (default) | $4.00 | $20.00 | Cache reads $0.20 · writes $5 |
| claude-opus-5 (compare) | $5.00 | $25.00 | Cache reads $0.50 · writes $6.25 |
| Opus 5.5 Fast mode | $8.00 | $40.00 | Up to 2.5× speed · Claude Code + Platform |
| Terminal-Bench 4.0 | — | 66.4% | Anthropic-reported vendor table |
Claude Sonnet 5.5 and Claude Haiku 5.5 are described as following “in the coming weeks” — treat that as roadmap, not a shipped SKU on this briefing date. Because Anthropic rates Opus 5.5 as comparable to Claude Mythos 5.1 in biology and cybersecurity, it deploys safeguards similar to those on Claude Fable 5.1, including the Life Sciences Verification Program and an expanding Cyber Verification Program.
What Alberta operators should do
For AI consulting Alberta and AI automation Alberta teams measuring agent dollars per task, the operator move is to re-price long agent loops against Opus 5 and against Fable 5.1 on the same harness: same task pack, same tools, then compare total token bill — not just list $/1M. Cache reads at $0.20 per 1M matter most on agentic coding where cached prefixes dominate cost.
Private AI security still applies when the prompt stack holds client tenders, plant credentials, or regulated notes: cite Anthropic’s Fable-class biology/cyber safeguards and verification programs as the vendor’s control story — do not invent Alberta customer case studies. Pair this briefing with Opcelerate’s AI consulting Alberta path and private AI security lane, and keep reading the city desk at The Super Intelligence Times. Foreign hosted API ≠ private Alberta inference.
The guardrail
Lower token cost is not permission to skip evaluation. Terminal-Bench 66.4% is Anthropic’s reported figure; Fable-class claims are Anthropic’s framing for “most work.” Do not treat Opus 5.5 as AGI or ASI. Budget Fast mode ($8/$40) only when the up-to-2.5× speed is worth the premium, and keep Sonnet/Haiku 5.5 off production plans until they ship.
claude-opus-5-5, $4/$20 per 1M with $0.20 cache reads, ~40% less than Opus 5 on typical workloads per Anthropic, Fast mode $8/$40 up to 2.5×. Attribute Terminal-Bench 66.4% to Anthropic only. Route sensitive repos through a private AI security review — hosted API ≠ on-prem.