Claude Opus 5.5 is Anthropic's model released on September 22, 2026: in the API it costs $4 per million input tokens and $20 per million output tokens, 20% less than Opus 5, and cache reads are 60% cheaper. There is no official free trial: Opus is included from the $20 Pro plan, and the cheapest way to test it is the API, where a test request costs a few cents. The model ID is claude-opus-5-5 (Anthropic announcement).
Your first Python request to claude-opus-5-5, the Messages API format and streaming are covered in Anthropic API Fundamentals. Agentic coding is covered in the Claude Code Course: Agentic Development on Real Projects. The first 2 lessons of every course are free, and full access comes with a 7-day trial, no card required.
Updated October 31, 2026: added the ways to try the model, batch pricing and the SDK changes that followed Sonnet 5.5 and Haiku 5.5. Figures are checked against the Claude pricing page and the Opus 5.5 documentation.
The biggest change is the cost of agentic work. Input and output tokens are 20% cheaper, cache reads are 60% cheaper, and Anthropic says cache reads make up the majority of agentic and coding costs. According to Anthropic, at default settings Opus 5.5 costs 40% less than Opus 5 on typical workloads and performs at the level of Claude Fable 5.1 on most work.
| Item | Opus 5.5 | Opus 5 | Change |
|---|---|---|---|
| Input tokens | $4 | $5 | −20% |
| Output tokens | $20 | $25 | −20% |
| Cache reads | $0.20 | $0.50 | −60% |
| Cache writes (5 min) | $5 | $6.25 | −20% |
Source: Pricing table in the Anthropic announcement.
Batch processing is half price: $2 for input and $10 for output. One-hour cache writes cost $8 per million tokens (Pricing section of the docs).
Anthropic itself notes that benchmark margins have become a less reliable guide to real-world differences, so test models on your own tasks.
Anthropic offers no official free trial of Claude Opus 5.5. The Claude Free plan includes Sonnet and Haiku, while Opus, Claude Code and Research start with Pro (claude.com/pricing, October 31, 2026). The pricing page doesn't promise free credits for new API accounts either. These are the official ways to get access and what each one costs.
| Option | Price | Best for |
|---|---|---|
| Pay-as-you-go API | $4 input and $20 output per 1M tokens; a test request costs a few cents | developers, and anyone who wants to test the model on 10 to 20 of their own tasks |
| Claude Pro | $20 a month, or $17 a month billed annually | daily work in the Claude apps and Claude Code |
| Claude Max and Ultra | $100 and $200 a month, with $100 and $200 in monthly API credits | people who outgrow Pro limits and also need the API |
| Claude Team | from $20 per seat a month billed annually, up to $500 in pooled API credits | teams |
| Cloud platforms | Amazon Bedrock, Google Cloud or Microsoft Foundry rates | companies with an existing cloud account and budget |
What an API test costs. A request with 2,000 input tokens and a 1,000-token answer costs about 3 cents: 2,000 × $4 / 1M = $0.008 plus 1,000 × $20 / 1M = $0.02. Thinking is billed as output tokens, so harder tasks cost more. To keep tests cheap, set effort to low or medium and cap max_tokens.
Sites offering "free Opus 5.5". These chats run through an intermediary: your prompts and files go to a third party, and you can't verify which model actually answers. For work data, use the official Claude apps, the API or the cloud platforms in the table.
The 7-day Academl trial unlocks courses and doesn't include access to Opus itself. In a week you can work through the API and Claude Code lessons and build your own test set, then check Opus 5.5 in the API for a couple of dollars.
Installing the SDK, setting ANTHROPIC_API_KEY and making your first call are covered in the free lesson First Steps: Installation and Your First Request. How to cut the bill with caching is in the Prompt Caching lesson of Claude API from A to Z.
Code written for Opus 5 can return a 400 error after you swap the model ID. Anthropic's docs list four changes:
thinking: disabled and a manual budget_tokens return an error; the effort parameter controls depth and defaults to medium (Opus 5 defaulted to high).tool_choice set to any or tool) returns an error. auto and none still work; for schema-valid JSON, Anthropic recommends auto with strict: true or structured outputs.computer_20251124 computer use tool is no longer accepted; use the computer_toolset_20260801 toolset.One more change alters the response shape without failing any request: the short notes the model writes between tool calls arrive in thinking blocks, and their text is empty by default. If your interface streamed those notes to users as progress updates, set the display parameter.
Support for claude-opus-5-5 arrived in Python SDK 1.8.0 on September 22, 2026. On October 21, Python SDK 2.0.0 shipped: it sends API version 2026-10-21 by default, where temperature, top_p and top_k return a 400 error, and it requires Python 3.10 or newer (release notes). Update with one command:
pip install -U anthropic
For a team, Opus 5.5 makes long agentic runs in Claude Code and multi-app automations noticeably cheaper. On Gray Swan's benchmark it ties Fable 5.1 for the lowest prompt injection success rate of any model tested, according to Anthropic.
A practical migration order:
medium and compare several levels on your own tasks, as the Opus 5.5 prompting guide recommends;max_tokens: thinking counts toward it even when the thinking text isn't returned;Prompts that worked on earlier models usually need a review after a model switch. A step-by-step way to build that skill is our 4-week prompt engineering plan.
Two more models followed Opus 5.5. Claude Sonnet 5.5 (claude-sonnet-5-5) launched on September 28 at $2 per million input and $10 per million output tokens. Claude Haiku 5.5 (claude-haiku-5-5), built for high-volume, latency-sensitive work, launched on October 7 at $0.10 and $0.50 for requests up to 100K tokens and $0.50 and $2.50 above that. Both have a 1M-token context window (release notes, API pricing). The top model in the lineup, Fable 5.1, costs $10 and $50.
Compare Claude with ChatGPT, Gemini, Grok and open-source models by task in Model comparison.
Anthropic offers no official free trial of Opus 5.5. The Claude Free plan includes Sonnet and Haiku; Opus starts with Pro at $20 a month. The cheapest way to test the model is the pay-as-you-go API, where a short test request costs a few cents.
$4 per million input tokens, $20 per million output tokens, $0.20 per million cache-read tokens and $5 per million 5-minute cache-write tokens. Batch processing is half price at $2 and $10. Fast mode costs $8 and $40 per million input and output tokens.
As of October 31, 2026, the Claude pricing page lists Sonnet and Haiku for the Free plan and Opus from Pro upward. Pro costs $20 a month, or $17 a month billed annually. Max ($100) and Ultra ($200) also include monthly API credits of $100 and $200.
Input and output tokens are 20% cheaper, cache reads 60% cheaper and output is more than 30% faster. Anthropic's benchmarks show gains in agentic coding and computer use. There are four breaking API changes: thinking can't be disabled, forced tool_choice errors, thinking blocks are tied to the model and conversation, and computer_20251124 is rejected.
Update the SDK with pip install -U anthropic and set model="claude-opus-5-5" in your request. Remove thinking: disabled and forced tool_choice, or the API returns a 400 error. Since October 2026, SDK 2.x sends API version 2026-10-21, where temperature, top_p and top_k also return 400.
Thinking in Opus 5.5 is always on and adaptive: the model decides how much to think. To get faster, cheaper answers, lower the effort parameter. It defaults to medium on Opus 5.5, while Opus 5 defaulted to high.
Sonnet 5.5 costs half as much: $2 and $10 per million tokens versus $4 and $20. Anthropic positions Opus 5.5 for long-running agentic coding and knowledge work. Run 10 to 20 of your own tasks on both models at the same effort and compare quality and cost.
Build your first agent on Opus 5.5: the 10 lessons of the Claude Code Course take you from setup to MCP and git workflows, and Anthropic API Fundamentals goes from your first request to vision. Start your 7-day trial, no card required, then Pro at $9 a month or $79 a year.