What is Claude Opus 4.5?
Opus 4.5 is the Opus model that cut Opus pricing to $5 / $25 per million tokens, built for coding, agents and computer use. It reads text and images, writes text, and works in a 200K-token context with up to 64K tokens of output. Its knowledge stops at May 2025. You switch thinking on and give it a token budget yourself, and an effort setting with three levels trims how many tokens it spends.
Anthropic now lists it as a legacy model and tells users to move to Opus 5.5, which costs less and reads five times as much context.
Is Claude Opus 4.5 still available?
Yes, for now: Anthropic lists Claude Opus 4.5 as a legacy model that is still active, but its retirement date is the soonest of any Opus model on the API.
- Status
- Active but legacy; Anthropic's model page tells users to consider moving to Opus 5.5.
- Where it runs
- The Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS, plus Claude Code.
- Retirement
- Anthropic's deprecation page gives a tentative retirement date, and Anthropic promises customers at least 60 days' notice before a model is retired.
- Replacement
- Opus 5.5, with a migration guide that has its own section for code written for Opus 4.5.
What changed in the Opus models after 4.5?
Mostly context and thinking: every later Opus reads 1M tokens instead of 200K, and newer ones decide for themselves how much to think.
- Context
- From Opus 4.6 on, Opus models take 1M tokens of context and write up to 128K tokens, against 200K and 64K here.
- Price
- Opus 4.6 through Opus 5 kept the $5 / $25 price; Opus 5.5 dropped it to $4 / $20.
- Thinking
- Opus 4.5 uses a thinking budget you set; from Opus 4.7 that setting returns an error and adaptive thinking takes over.
- Knowledge
- Opus 4.5 knows the world up to May 2025; Opus 5.5 knows it up to June 2026.
Who is Claude Opus 4.5 best for?
Claude Opus 4.5 is best for teams whose apps and evals are already tuned to it, and for coding and agent work at $5 / $25 per million tokens while they plan a move to a newer Opus.
Apps already built on Opus 4.5
Teams whose prompts and evals were tuned against this model ID can keep the same behavior while they plan a move to a newer Opus.
Coding and agent work
Anthropic built it for coding, agents and computer use, and it led Sonnet 4.5 in 7 of 8 languages on SWE-bench Multilingual.
Keeping token counts down
The effort setting trades depth for cost; at medium effort it matched Sonnet 4.5's best SWE-bench Verified score with 76% fewer output tokens.
Multi-agent systems
Anthropic reports it handles a team of subagents well, which suits setups where one model plans and others carry out the steps.
What's included with Claude Opus 4.5?
Three effort levels, extended thinking with a budget you set, image input, computer use, prompt caching with half-price batch jobs, and five platforms.
- Effort control
- Low, medium and high effort levels trade answer depth against token spend; high is the default.
- Extended thinking
- You turn thinking on and set its token budget per request, where Opus 5.5 always thinks and sizes its own reasoning.
- Image input
- It reads screenshots, charts and documents as images and scored 80.7% on MMMU in Anthropic's launch table.
- Computer use
- It drives a desktop through the computer_20251124 tool under a beta header, and scored 66.3% on OSWorld at launch.
- Prompt caching and batch jobs
- Cache reads cost $0.50 per million tokens, and the Batch API halves input and output prices.
- Five platforms
- It runs on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS, and Claude Code still lists it.
How much does Claude Opus 4.5 cost?
Claude Opus 4.5 costs $5 per million input tokens and $25 per million output tokens on the Claude API, and half that through the Batch API.
- Cache reads $0.50, 5-minute cache writes $6.25, 1-hour writes $10
- Batch API at half price: $2.50 / $12.50
- Also on Amazon Bedrock and Google Cloud, billed by them
Anthropic now lists Opus 4.5 under legacy models. It is still active on the API as claude-opus-4-5-20251101, but its deprecation table gives a tentative retirement date. Opus 4.6, 4.7, 4.8 and Opus 5 cost the same $5 / $25, and Opus 5.5 costs less at $4 / $20. Regional and multi-region endpoints on Amazon Bedrock and Google Cloud cost 10% more than global ones. Prices are in US dollars. Full pricing on claude.com. We checked these prices on .
How do readers rate Claude Opus 4.5?
No one has rated Claude Opus 4.5 yet. Be the first.
Sign in to rate. One rating per account.
Claude Opus 4.5 pros and cons
Developers rate it for big refactors and long coding sessions, but its 200K context, fast-draining subscription limits and near retirement count against it.
Pros
- Developers on Hacker News trust it with heavy refactors and large, poorly documented codebases, where it finds its way between files and tools.
- Its cost per finished task can beat its list price: one coding tool team measured $1.30 per thread against $1.83 for Sonnet 4.5, because it thinks less and makes fewer wrong turns.
- Daily users in Claude Code, Cursor and GitHub Copilot called it the best model they had tried for real coding work, and one said it could be left alone on a task.
- It scored 80.9% on SWE-bench Verified in Anthropic's launch table, ahead of Sonnet 4.5 at 77.2% and Opus 4.1 at 74.5%.
Cons
- Its 200K-token context is a fifth of what Opus 4.6 and every later Opus model take, and users say it runs out of room on large tasks.
- On Claude subscriptions, heavy Opus 4.5 use, especially with subagents, drained five-hour and weekly limits fast for several users.
- Developers working on low-level, niche or brand-new code say it gets lost outside well-trodden patterns and burns tokens going in circles.
- It is a legacy model with the soonest retirement date of any active Opus model, and moving to Opus 5.5 means code changes: no assistant prefills, adaptive thinking instead of a token budget, and new parameter names.
What are the best Claude Opus 4.5 alternatives?
Opus 5.5 as the named replacement, Opus 4.8 to keep the same price with a 1M context, and Sonnet 5 for the lowest cost.
Claude Opus 5.5PaidChoose it when you can re-test your prompts; it is Anthropic's named replacement, costs less at $4 / $20 per million tokens, and reads 1M tokens with knowledge up to June 2026.
Claude Opus 4.8PaidChoose it when you want to stay at the same $5 / $25 price but need a 1M-token context, 128K tokens of output and a later retirement date.
Claude Sonnet 5FreeChoose it when cost matters most; at $2 / $10 per million tokens it is less than half the price, with a 1M-token context and knowledge up to January 2026.
Questions about Claude Opus 4.5
What is Claude Opus 4.5?
Claude Opus 4.5 is Anthropic's legacy Opus model for coding, agents and computer use, still served on the API at $5 / $25 per million tokens; Anthropic now points users to Opus 5.5.
Who is Claude Opus 4.5 for?
Claude Opus 4.5 is best for teams whose apps and evals are already tuned to it, and for coding and agent work at $5 / $25 per million tokens while they plan a move to a newer Opus.
Is Claude Opus 4.5 free?
No. Claude Opus 4.5 is a paid tool, from $5 / $25 per million input / output tokens.
How much does Claude Opus 4.5 cost?
Claude Opus 4.5's paid plans are Claude API at $5 / $25 per million input / output tokens.
What are the best Claude Opus 4.5 alternatives?
The closest alternatives on Recaply are Claude Opus 5.5, Claude Opus 4.8 and Claude Sonnet 5.
Where can I try Claude Opus 4.5?
On its official website, anthropic.com.
Is Claude Opus 4.5 cheaper than Claude Opus 5.5?
No. Claude Opus 4.5 costs $5 per million input tokens and $25 per million output tokens, while Opus 5.5 costs $4 and $20 on the same pricing page. Opus 4.6, 4.7, 4.8 and Opus 5 cost the same as Opus 4.5.
How big is Claude Opus 4.5's context window?
Claude Opus 4.5 takes up to 200K tokens of input, roughly 150,000 words, and writes up to 64K tokens per response. Opus 4.6 and every later Opus model take 1M tokens and write up to 128K.
Claude Opus 4.5 vs Sonnet 4.5: which is better?
Opus 4.5 scored 80.9% on SWE-bench Verified against 77.2% for Sonnet 4.5 in Anthropic's launch table, and at medium effort it matched Sonnet 4.5's best score with 76% fewer output tokens. Sonnet 4.5 is cheaper at $3 / $15, and both are now legacy models.
Can I still use Claude Opus 4.5 in Claude Code?
Yes. Anthropic's Claude Code help page still lists Opus 4.5 as a supported model. Pick it from the /model menu, or start a session with claude --model claude-opus-4-5-20251101.
How do I move from Claude Opus 4.5 to Opus 5.5?
Change the model ID to claude-opus-5-5, remove assistant-message prefills, replace the thinking token budget with adaptive thinking and an effort level, drop the old beta headers, and move output_format to output_config.format. Anthropic's Opus 5.5 migration guide has a section for Opus 4.5.
Our verdict
Keep Claude Opus 4.5 only while you move off it; new work belongs on Opus 5.5.
It still runs at $5 / $25 per million tokens, but Opus 5.5 costs less, reads five times as much context and knows about a year more, and Anthropic already lists Opus 4.5 as legacy.
- Skip it for
- New projects, and anyone who needs more than 200K tokens of context.
- Check first
- Check its retirement date on Anthropic's model deprecations page and plan for the migration changes.
Claude Opus 4.7
Claude Fable 5
Claude Fable 5.1
Claude Sonnet 4.6
