Anthropic Releases Claude Opus 5.5, Claims 40% Lower Running Costs Than Opus 5

Anthropic released Claude Opus 5.5 on September 22, 2026. According to the company, the model leads in agentic coding and knowledge work and costs 40% less to run than Claude Opus 5 on typical workloads.

What Anthropic has said

The only details available from Anthropic's announcements index are a one-sentence summary and a date. Both claims below are Anthropic's own and have not been independently verified:

• Capability: Opus 5.5 "leads in agentic coding and knowledge work."
• Cost: It costs 40% less to run than Opus 5 "on typical workloads."

What has not been disclosed

The summary does not include the following, so we are not reporting figures for them:

• Context window size: not yet disclosed
• Pricing per 1M input and output tokens: not yet disclosed
• Benchmark scores (including agentic coding evaluations): not yet disclosed
• Parameter count and training cutoff date: not yet disclosed

The 40% figure is also ambiguous. Anthropic's wording says "costs 40% less to run," which could reflect lower list prices, fewer tokens used per task, or both. The phrase "typical workloads" means the saving will vary with usage. Teams should measure against their own traffic rather than assume a flat 40% reduction.

The phrase "leads" is also unqualified in the summary. It does not say whether the comparison is against other Anthropic models, competitors, or both. Anthropic's September 15 announcement of Claude Mythos and Fable 5.1 described those models as "the world's most advanced models for coding and knowledge work," so the Opus 5.5 claim likely refers to a specific comparison set.

Release context

Opus 5.5 is the first of a cluster of 5.5-generation releases:

• Sep 22, 2026: Claude Opus 5.5
• Sep 28, 2026: Claude Sonnet 5.5, which Anthropic says runs 30%+ faster than Sonnet 5 and costs up to 30% less for most work
• Oct 7, 2026: Claude Haiku 5.5, described as Anthropic's fastest, most capable small model, aimed at summarization, subagents and browser use

The Opus 5.5 announcement came five days after Anthropic launched its Life Sciences Verification Program, which gives qualifying professionals access to Claude Mythos, Opus and Sonnet models with safeguards tuned for biology-related work.

What this means

The headline claim is about cost, not just capability. Anthropic is pairing a top-tier model with a reduction in run cost, which matters most for agentic coding workloads that consume many tokens across long multi-step tasks. If the 40% figure holds on real workloads, it lowers the cost of running Opus-class models in production agents.

The claim cannot yet be evaluated, though. Without published per-token prices, a context window and benchmark results, developers cannot tell whether the saving comes from pricing or efficiency, or how Opus 5.5 compares with Mythos, Fable 5.1 or competing frontier models. We will update this article when Anthropic publishes the full model card and API pricing.