Opus 5.5 Release Date: Anthropic Ships Its New Flagship

Anthropic released Claude Opus 5.5 on September 22, 2026, the first model in a new Claude 5.5 family. Anthropic says it performs at the level of Fable 5.1 on most work and, at default settings, costs 40% less to run than Opus 5 on typical workloads. Input and output tokens are priced at $4 and $20 per million, a 20% cut from Opus 5, and the model is available across every major platform today.

The short version for anyone who was tracking the leaks: the release is real, the pricing is lower than Opus 5, and the "version jump" rumor was wrong. This is the start of a 5.5 line, not a skipped release.

When did Opus 5.5 come out

Opus 5.5 launched on September 22, 2026, the same day Anthropic published its announcement. Anthropic announced immediate availability across its supported platforms. The model is available on the Claude API, Amazon Bedrock and Claude Platform on AWS, Google Cloud, and Microsoft Foundry (Azure). Developers can call it with the model string claude-opus-5-5, and Anthropic's platform release notes list a 1 million-token default context window, 128,000 maximum output tokens, and always-on adaptive thinking.

Anthropic also confirmed the family will grow. Claude Sonnet 5.5 and Claude Haiku 5.5 are set to follow in the coming weeks, carrying many of the same gains in performance, efficiency, and safety.

What Opus 5.5 does better

Anthropic positions Opus 5.5 as its leading model for agentic coding, computer use, and knowledge work, while cautioning that benchmark margins at this level are becoming less reliable guides to real-world differences. Its headline advantage is doing that work for less. Anthropic reports that an early tester completed a 680,000-line code migration in under a day, and that another audited and fixed a 200,000-line codebase in under three hours, work that took Opus 5 over 20 hours and 2.5 times as many tokens. These are tester reports rather than independently audited benchmarks.

The efficiency story is the throughline. Opus 5.5 costs less per token than Opus 5 and uses fewer tokens per task, which nets out to the 40% drop in cost. Anthropic also reports it generates output more than 30% faster than Opus 5. On the FrontierCode benchmark, Anthropic says Opus 5.5 at default effort beats GPT-6 Astra at roughly a fifth of the cost per task. That comparison is not a clean same-operator result, though: Anthropic reports its own Opus figure and cites OpenAI's reported figure for Astra, so the two runs are not directly comparable.

Table 1: Confirmed Opus 5.5 pricing versus Opus 5, per million tokens

Pricing as of September 2026, confirmed via Anthropic's Opus 5.5 announcement. Fast mode runs at $8 input and $40 output per million, at up to 2.5x speed.

How the leaks compared to reality

The rumor cluster got the direction right and some specifics wrong. The leaked $4 input and $20 output prices matched the launch exactly. Several leaks framed the release as Anthropic "skipping a version to fight GPT-6." Anthropic's own announcement frames it differently, as the first entry in a new Claude 5.5 family rather than a jump past a numbered release. On the context window, Anthropic's platform documentation confirms a 1 million-token default and 128,000 maximum output tokens, though the launch announcement itself foregrounds performance, cost, safety, and availability rather than those specifications.

This is the usual pattern with pre-launch leaks. A few details land, the framing distorts, and the only reliable version arrives when the vendor publishes it. Anthropic's announcement is now the primary source, and it supersedes every earlier screenshot.

Safety and the pacing question

Anthropic reports that Opus 5.5 is the strongest-performing model it has recorded on its automated behavioral audit, the alignment suite it runs across nearly 2,000 scenarios. By Anthropic's account, the model is less likely than recent releases to take hard-to-reverse actions or act outside its given boundaries, and more resistant to prompt injection than Opus 5.

The release is also Anthropic's first since its CEO called for pacing the frontier, and the safeguards reflect that. Because Anthropic says Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, it launches with safeguards similar to those used for Fable 5.1. When those safeguards intervene, most cybersecurity tasks are handled by Opus 4.8, while biology and frontier-LLM-development tasks are handled by Opus 5. Vetted organizations can apply to Anthropic's Life Sciences Verification Program to use the model for biology research.

What it means for builders

For anyone building on Claude, Opus 5.5 lowers the cost of the most demanding work rather than adding a premium tier on top. The same coding and research jobs now run cheaper and faster, and the model is the new default choice for complex, long-running tasks. Pro, Max, Team, and seat-based Enterprise plans also get higher five-hour usage limits alongside the price drop.

Opus 5.5 is now available on Emergent, so you can build with the new flagship the same day it shipped. You can start building on Emergent with the Universal LLM Key, which gives single-credential access with unified billing across supported Claude, GPT, and Gemini models. That single credential means your project runs on current models rather than last quarter's.

The takeaway is simple. Opus 5.5 is out, it is cheaper to run than Opus 5, and it sets a new bar for agentic coding and knowledge work. The rumors are settled because the model is now on the table.