In an industry that usually equates "better" with "bigger and more expensive," Anthropic just released a model whose whole pitch is the opposite: do what the flagship does, charge a lot less, and don't make the user wait around for it.
A Flagship-Grade Model at Mid-Tier Prices
On September 22, Anthropic launched Claude Opus 5.5, the first entry in its new 5.5 family, priced at $4/$20 per million tokens — a 20% cut from Opus 5 — with cache reads down 60% to $0.20 per MTok. Anthropic says it matches or beats its larger Fable model on a range of benchmarks and informal tasks, while cutting typical workload costs by 40% and boosting output speed over 30%.
The model ships with a 1-million-token context window by default, up to 128,000 tokens of output, and "always-on" adaptive thinking tuned by effort level rather than a simple on/off toggle. In practical terms, Anthropic points to a legacy HAProxy port from C to Rust finishing in 9.5 hours versus 12 for Fable 5.1, and a 200,000-line code audit dropping from over 20 hours to under three. It's live now on AWS, Google Cloud, and Azure, with Sonnet 5.5 and Haiku 5.5 said to be coming within weeks.
The Real Story Is the Price Curve
Model quality headlines are fun, but the number that actually changes what businesses can build is the cost-per-task, not the leaderboard rank. When a model this capable gets 40% cheaper to run, tasks that were "nice to have but too expensive to automate" quietly become "why are we still doing this by hand."
There's also a quieter shift here worth noting: Anthropic tuned Opus 5.5 to front-load key information and cut jargon in its responses. That's a small thing on paper, but for any business embedding AI into customer-facing workflows, an assistant that communicates more plainly by default is less work to prompt-engineer around later.
Cheaper, faster, and better at getting to the point — that's a rare trifecta, and it's going to reshape what "worth automating" means for a lot of businesses this quarter.
If a price drop like this just moved one of your team's "someday" AI ideas into "actually feasible now" territory, that's exactly the kind of integration work the WTK team likes to dig into — reach out and let's scope it.
Source: TechCrunch