Claude Opus 5.5: What's New and How to Try It

Anthropic just shipped its strongest model

Claude Opus 5.5 launched on September 22, 2026 — and the headline numbers are hard to ignore. It performs at the level of Claude Fable 5.1 on most tasks, costs 40% less to run than Opus 5, and generates output over 30% faster.

This is the first release since Anthropic CEO Dario Amodei publicly called for "pacing the frontier," so the model was tested by external evaluators including Frontier Design and METR before going live.

Here is what actually matters for developers and everyday users, how the benchmarks compare, and every way to try it right now.

What changed from Opus 5

Performance

Opus 5.5 is built for long, complex tasks — the kind where previous models would lose context or produce inconsistent output. Anthropic's own examples:

An early tester completed a 680,000-line code migration in under a day — work that would have taken an engineering team weeks.
When asked to cut load times across every page of a web app, Opus 5.5 succeeded 39 out of 40 times, while Opus 5 made smaller improvements that also broke app behavior.
In an internal test, Opus 5.5 translated HAProxy (a widely-used C load balancer) to Rust in 9.5 hours vs. 12 hours for Fable 5.1, at 51% less cost.

Better communication

Previous Claude models sometimes buried the important information or added unnecessary caveats. Opus 5.5 has been specifically trained to write more clearly — putting the key point first and reducing conversational drag.

As one early tester described it: "it writes the way I do."

This matters for practical work. When the model communicates clearly, you spend less time re-reading and more time acting on the output.

Safety improvements

Opus 5.5 achieves the highest scores on Anthropic's automated behavioral audit — their most comprehensive alignment test. Key improvements:

Less likely to take hard-to-reverse actions or act outside given boundaries
More resistant to prompt injection than Opus 5
Broader alignment testing covering longer tasks, impossible tasks, and real-incident scenarios

Biology and cybersecurity safeguards from Fable 5.1 carry over. Vetted researchers can apply to Anthropic's Life Sciences and Cyber Verification programs for unrestricted access.

Benchmark results

Benchmark	Opus 5.5	Fable 5.1	Opus 5	GPT-6 Astra
Terminal-Bench 4.0 (agentic coding)	66.4%	55.8%	52.3%	57.9%
FrontierCode v1.1 (agentic coding)	54.4%	50.3%	48.0%	53.3%
CursorBench 4.0 (agentic coding)	57.8%	51.8%	46.6%	—
GDPval-AA v2.1 (knowledge work)	1846	1735	1708	1542
Humanity's Last Exam	67.7%	65.6%	63.6%	57.2%
OSWorld 2.0 (computer use)	81.8%	80.7%	74.0%	—

Anthropic notes that at these capability levels, benchmark margins are a less reliable guide to real-world differences. The gap between Opus 5.5 and Fable 5.1 is narrower than the numbers suggest in daily use.

Pricing — 40% cheaper than Opus 5

The biggest practical improvement is cost. Token prices dropped across the board, and Opus 5.5 uses fewer tokens per task.

	Opus 5.5	Opus 5
Input tokens	$4/M	$5/M
Output tokens	$20/M	$25/M
Cache reads	$0.20/M	$0.50/M
Cache writes	$5/M	$6.25/M

Cache reads are where the big savings happen — 60% cheaper. Since cached reads make up the majority of costs in agentic and coding workflows, the real-world savings are significant.

A Fast mode is also available at $8/M input and $40/M output for time-sensitive tasks, with up to 2.5x speed.

How to try Claude Opus 5.5

Free options

Claude.ai free tier. Go to claude.ai and create an account. The free tier gives you access to Claude's models with usage limits. Opus 5.5 may be available with limited daily messages depending on current rollout.

API credits for new accounts. When you sign up for a new Anthropic Console developer account, you receive starter credits to test the API — enough to experiment with Opus 5.5 on real tasks.

Third-party platforms. Services like OpenRouter and some cloud providers offer free-tier credits that include Claude models. Check current availability, as free allocations change.

Paid options

Claude Pro ($20/month). The most practical way to use Opus 5.5 regularly. Includes higher usage limits and access to Claude Code for terminal-based agentic coding. Anthropic has increased five-hour usage limits and added a rate limit reset you can save and use when you choose.

Claude Max (from $100/month). 5x or 20x the capacity of Pro. Built for heavy agentic workloads and extended coding sessions.

Claude Team and Enterprise. Seat-based plans for organizations. All plans received increased usage limits with this release.

API (pay-per-token). Use model ID claude-opus-5-5 in the Anthropic API. Best for developers building applications on top of Claude. Available through the Anthropic API directly, Amazon Bedrock, Google Cloud, and Microsoft Foundry.

How to use it in Claude Code

If you use Claude Code for terminal-based development, Opus 5.5 is available immediately. The model is designed for the kind of long-running agentic tasks that Claude Code excels at — multi-file refactoring, codebase migrations, and autonomous test-and-fix loops.

Key things to know:

Adaptive thinking is always on. You control depth with the effort parameter (low, medium, high, max). Default is medium.
Thinking cannot be disabled. This is a breaking change from previous models. The model always shows its reasoning.
Fast mode works in Claude Code. Use claude-opus-5-5-fast for speed-critical tasks.
1M token context window with 128K max output — large enough to hold entire codebases in context.

How it compares to ChatGPT and other models

If you're deciding between Claude and ChatGPT, Opus 5.5 makes the strongest case yet for Claude in coding and knowledge work. It leads GPT-6 Astra on most agentic coding benchmarks at significantly lower cost per task.

For general-purpose everyday use — web browsing, image generation, voice mode — ChatGPT still has broader integrations. The choice depends on your workflow.

For AI coding tools specifically, Opus 5.5 powering Claude Code is the current state of the art for complex, multi-file development work.

What's coming next

Anthropic has announced that Claude Sonnet 5.5 and Haiku 5.5 will follow in the coming weeks, bringing the same improvements in performance, efficiency, and safety to smaller and faster models.

The pattern is clear: each generation gets significantly cheaper and faster while matching or exceeding the previous flagship's capabilities. For developers and teams already using Claude, Opus 5.5 is a direct upgrade. For everyone else, the free tier and starter API credits make it easy to test whether it fits your workflow.