Claude Opus models: Opus 5.5, Opus 5 and older versions

Every Claude Opus model you can still use

Here are the Claude Opus versions currently running on the API, with their model IDs, release dates, pricing and context limits.

• Opus 5.5 (claude-opus-5-5) is the latest release, while Opus 5, 4.8, 4.7, 4.6 and 4.5 are legacy models that still work on the API but get no further updates.
• Claude Opus 4.1 (claude-opus-4-1-20250805, retired August 5, 2026) and Claude Opus 4 (claude-opus-4-20250514, retired June 15, 2026) are shut down on the Claude API, so requests to them fail and Anthropic points callers to claude-opus-4-8 instead.
• Anthropic gives at least 60 days notice before retiring any publicly released model.

Claude Opus 5.5 specs and pricing

The table below covers the technical limits, model IDs and token rates for Claude Opus 5.5.

Anthropic says Opus 5.5 performs at the level of Claude Fable 5.1 on most work. On typical workloads billed by token, it costs 40% less to run than Claude Opus 5 and writes output more than 30% faster.

If you don't need immediate responses, the Message Batches API cuts token costs by 50% and supports up to 300K output tokens when you pass the output-300k-2026-03-24 beta header. US-only inference is billed at 1.1x the base input and output rates, and prompt caching requires a minimum prompt length of 512 tokens; see our Claude API pricing guide for a full breakdown.

Opus 5.5 benchmarks vs Fable 5.1 and Opus 5

These are the launch benchmark scores Anthropic published comparing Opus 5.5, Fable 5.1 and Opus 5. The runs use adaptive thinking at max effort unless noted (Terminal-Bench 4.0 ran at xhigh effort).

What changed between Opus versions

• Opus 5.5 (September 22, 2026): Delivers Fable 5.1-level results on most work at $4 input and $20 output per million tokens, with cache reads 60% cheaper than Opus 5 at $0.20. Adaptive thinking is always on and can't be disabled, and the default effort parameter is medium.
• Opus 5 (July 24, 2026): Came close to Claude Fable 5's intelligence at half the price, keeping the $5 / $25 per million token rate of Opus 4.8 with a high default effort. It launched as the default model on Claude Max and the strongest option on Claude Pro.
• Opus 4.8 (May 28, 2026): Added stronger coding, agentic and professional performance over 4.7, with better consistency and autonomy to keep working through long-running tasks.
• Opus 4.7 (April 16, 2026): Improved coding, vision and complex multi-step work. It also introduced Claude's current tokenizer, which fits about 555k words into 1M tokens instead of the roughly 750k words on earlier models.
• Opus 4.6 (February 5, 2026): Raised the context window to 1M tokens and max output to 128K, up from the 200K context window and 64K max output on Opus 4.5. It also moved to adaptive thinking and deprecated manual extended thinking.

Moving from Opus 5 to Opus 5.5

If you're upgrading existing code from claude-opus-5 to claude-opus-5-5, check for these four breaking changes before switching the model ID:

1. Adaptive thinking is always on and can't be disabled.
2. Forced tool use is no longer supported and returns an error.
3. Thinking blocks are tied to both the specific model and the conversation.
4. On the Claude API and Claude Models on Google Cloud (Google managed), the older computer_20251124 computer use tool is rejected.

Which Claude plans include Opus

• Opus 5.5 is included on Claude Pro, Max, Team and Enterprise subscriptions.
• Claude's Free plan doesn't include access to any Opus model.
• Along with the Opus 5.5 release, Anthropic raised five-hour Claude usage limits across Pro, Max, Team and seat-based Enterprise plans, and added a rate limit reset that subscribers can save and use whenever they want.
• For API workloads outside the Claude Platform, developers can call Opus 5.5 through Amazon Web Services (anthropic.claude-opus-5-5 on Bedrock), Google Cloud and Microsoft Foundry.

If you're still deciding whether your workload needs Opus at all, see our Claude Sonnet vs Opus comparison.

Use Opus from your Claude subscription in code

Plan2API lets you connect a Claude Pro or Max account and call claude/claude-opus-5-5, claude/claude-opus-5, claude/claude-opus-4-8 and claude/claude-opus-4-7 (or with the anthropic/ prefix and dotted names like claude-opus-5.5) through a standard OpenAI Chat Completions endpoint at https://api.plan2api.com/v1. You pass your p2a_ key in the Authorization: Bearer header, requests count against your Claude subscription's own usage limits, and there's no per-token API bill.

The Plan2API Free plan costs $0 with no card required and includes 1 connected subscription and 50 chat requests a day, while the Pro plan is $12 a month (or $99 a year) for up to 3 connected subscriptions and no daily request cap; see the Claude subscription API page for setup details. A Python example using the OpenAI SDK follows below.

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.plan2api.com/v1",
    api_key=os.environ["PLAN2API_KEY"],
)

response = client.chat.completions.create(
    model="claude/claude-opus-5-5",
    messages=[{"role": "user", "content": "Review this function for race conditions."}],
)
print(response.choices[0].message.content)