Anthropic shipped Claude Opus 5.5 on September 22. Most of what's circulating about it online is secondhand, blog posts and pricing trackers citing Anthropic's launch materials rather than the source itself.
I went through the actual numbers and cross-checked them. Here's what multiple independent sources agree on, versus the claims that only show up once.
See current Claude plan pricing on Anthropic's official site
Affiliate link — we may earn a commission at no extra cost to you.
Key Takeaways
- List pricing reportedly dropped to $4 per million input tokens and $20 per million output, down from Opus 5's $5 and $25, a roughly 20% cut on paper.
- Output speed is reported as more than 30% faster than Opus 5, with adaptive "thinking" now defaulting to medium effort instead of high.
- Anthropic claims typical workloads cost about 40% less overall, but independent testing found that claim doesn't hold at maximum effort settings, so your actual savings depend heavily on how you configure it.
- Context window is 1M tokens with a 128K token maximum output, unchanged in scale from what Opus 5 offered.
- Sonnet 5.5 and Haiku 5.5 are reportedly coming later, not available at this Opus 5.5 launch.
01.The Pricing Change
The most consistently reported number across trackers: input dropped from $5 to $4 per million tokens, output from $25 to $20 per million. Roughly a 20% list-price cut.
Cached reads reportedly fell to $0.20 per million tokens, a 60% reduction from $0.50 by one source's math. One tracker also lists a batch-processing rate of $2/$10 and a "fast mode" at $8/$40, but that breakdown only appears in a single source. I'd treat it as unconfirmed until it's echoed elsewhere or shows up on Anthropic's own pricing page.
02.What Changed Under the Hood
Opus 5.5 reportedly uses adaptive thinking that's always on, but now defaults to medium effort rather than the high default Opus 5 shipped with. The same prompt likely burns fewer reasoning tokens by default, unless you explicitly raise the effort setting. Output is reported as more than 30% faster.
On benchmarks, Anthropic's own comparison table reportedly places Opus 5.5 at or above both Opus 5 and Claude Fable 5.1. One secondary source cites Terminal-Bench 4.0 at 66.4% versus Opus 5's 52.3%, and a Terminal-Bench-Science gap of 58.7% versus 29.0%. A meaningful jump, if accurate.
The important caveat, repeated by more than one source: these are Anthropic-reported numbers, largely run at maximum effort. A flattering condition that may not reflect your default usage.
03.The Cost Savings Claim, Checked
Anthropic's stated claim: typical workloads cost about 40% less than on Opus 5. That doesn't survive scrutiny uniformly.
One independent test found no savings at all running at maximum effort. A separate analysis argues the straight rate cut only accounts for about 24 of those 40 points, implying the rest comes from the lower default effort setting using fewer tokens per response, not the price change alone.
The practical read: if you're running Opus 5.5 at default settings on typical tasks, a real cost reduction is plausible. If you're forcing maximum effort for every call, don't assume the 40% figure applies to you.
04.Should You Switch From Opus 5?
If you're already on Claude and primarily doing long-document work or coding, the combination of a real list-price cut and faster output makes upgrading low-risk. For a broader look at how Claude compares against ChatGPT for everyday work rather than just this specific release, our Claude AI Review still holds up on the fundamentals, Opus 5.5 sharpens the strengths that review already identified rather than changing the overall competitive picture.
Q: Is Claude Opus 5.5 actually 40% cheaper to run than Opus 5? A: Only under specific conditions. The list price itself dropped about 20%, and the rest of the claimed 40% savings appears to come from the new medium-effort default using fewer tokens, not a pure price cut. At maximum effort settings, independent testing found no savings at all.
Q: What's the difference between Claude Opus 5.5 and the upcoming Sonnet 5.5 or Haiku 5.5? A: Opus 5.5 is Anthropic's top-tier model, released first. Sonnet 5.5 and Haiku 5.5 are reportedly coming later as smaller, cheaper tiers in the same generation, but weren't available at this Opus 5.5 launch, so direct comparison isn't possible yet.
Q: Where can I verify these pricing and benchmark numbers myself? A: Check Anthropic's official pricing page and launch announcement directly. This article draws on multiple third-party trackers that cite Anthropic's materials secondhand, and where those sources disagreed, we've flagged it rather than picking one.
05.The Bottom Line
Claude Opus 5.5 is a real, confirmed upgrade, lower list pricing, faster output, and benchmark gains that multiple sources corroborate directionally even if the exact numbers vary. The 40%-cheaper headline claim needs a caveat: it depends on your effort settings, not just the new price list. If you're already a Claude user, this is a low-risk upgrade; if you're choosing between providers for the first time, it strengthens Claude's position without fundamentally changing the tradeoffs in our existing comparison coverage.
For everything else that shipped around the same time, see our AI News Roundup: Every Major Model Release in October 2026.
