Modelpeers
Claude Opus 5 charactervsClaude Opus 5.5 character

Claude Opus 5 vs Claude Opus 5.5: Should you upgrade?

Facts checked Oct 22, 2026. Every number links to its source.

Summary

Derived from sourced facts, not from builder opinions
  • Opus 5.5 costs 20% less: $4 input and $20 output per million tokens, against $5 and $25 for Opus 5.
  • With Claude Code as the agent, Opus 5.5 scored 64.85% on Terminal-Bench 4.0 at max effort. Opus 5 scored 53.94% at its best setting, xhigh.
  • It is not a drop-in swap: on Opus 5.5, turning thinking off or forcing a tool choice returns an error.
  • Opus 5 stays available as a legacy model, with retirement no sooner than July 24, 2027.

Choose Claude Opus 5 if

  • Your app turns thinking off or forces a specific tool, and you cannot change that yet.
  • You have tuned prompts on Opus 5 and cannot retest before mid-2027.
  • You rely on Opus 5's high default effort and cannot retest output quality and length yet.

Choose Claude Opus 5.5 if

  • You want the lower price and the higher Terminal-Bench 4.0 score.
  • You are starting new work or can retest your prompts now.
  • You want the newer knowledge cutoff (June 2026 against May 2026).
  • You want a model Anthropic still updates. Opus 5 is legacy, and its retirement can come with 60 days' notice.
Claude Opus 5
Claude Opus 5.5

Follow both to get an email when either one changes its price, limits or behavior.

Best by task

Agentic coding (Terminal-Bench 4.0 official board, both in Claude Code)

Claude Opus 5.5

64.85% at max effort against 53.94% at xhigh, 330 trials each. The 95% confidence intervals do not overlap.

Web development (Arena WebDev)

Claude Opus 5.5

Rank 1 (1807) for Opus 5.5 (Max) against rank 8 (1690) for Opus 5 (Max).

Overall capability (Epoch Capabilities Index)

Tied

168.83 against 163.94. Opus 5.5 is ahead, but the 95% confidence intervals overlap (165.25 to 173.95 and 161.12 to 168.24).

Chat preference (Arena text)

Tied

Rank 2 (1516) for Opus 5.5 (High) against rank 4 (1508) for Opus 5 (Max). The confidence ranges overlap.

Cost per test on Terminal-Bench 4.0 (Vals AI, Mini-SWE-agent harness, API prices)

Claude Opus 5.5

$13.20 per test against $18.60, and Opus 5.5 also solved more tasks (65.15% against 53.53%).

Cost

One request in an agent loop that rereads a repository, so most input is cached. A full task makes many such requests. Illustrative, not measured.

Claude Opus 5

$0.170 per request

$170.00 a month for 1,000 requests

Claude Opus 5.5

$0.124 per request

$124.00 a month for 1,000 requests

Claude Opus 5.5 costs 27% less for this workload at list prices.

Estimate per request: (input × input rate + cached input × cached rate + output × output rate) ÷ 1,000,000, at list prices on the cited pricing pages. It is not a measured task cost: models use different numbers of tokens, including reasoning tokens, for the same task.

Side by side

Claude Opus 5Claude Opus 5.5
MakerAnthropic
Anthropic
Model IDclaude-opus-5 (Bedrock: anthropic.claude-opus-5)
claude-opus-5-5 (Bedrock: anthropic.claude-opus-5-5)
Released2026-07-24
2026-09-22
StatusActive (legacy)
Not published
API price per 1M tokens$5 input / $25 output · 5-minute cache write $6.25, 1-hour cache write $10, cache read $0.50 · Batch 50% off
$4 input / $20 output · 5-minute cache write $5, 1-hour cache write $8, cache read $0.20 · Batch 50% off
Context window1M tokens
1M tokens
Max output128K tokens (300K on Batch API with beta header)
128K tokens (300K on Batch API with beta header)
Knowledge cutoffMay 2026 (reliable knowledge and training data)
June 2026 (reliable knowledge and training data)
Reasoning settingshigh · adaptive thinking (can be turned off, unlike Opus 5.5)
medium · adaptive thinking always on
Where to use itClaude API, Amazon Bedrock, Google Cloud (Google managed and Anthropic managed), Microsoft Foundry, Claude Platform on AWS
Claude API, Amazon Bedrock, Google Cloud (Google managed and Anthropic managed), Microsoft Foundry, Claude Platform on AWS · also Claude Code
Retires no sooner than2027-07-24
2027-09-22
Fast mode priceNot published$8 input / $40 output (up to 2.5x speed, Claude Code and Claude Platform)
Quality tracking (MarginLab)Not publishedClaude Code with Opus 5.5/high on a SWE-Bench-Pro subset: 77% latest day (50 tasks), 76% 7-day, 81% 30-day · baseline still being collected, detection paused (last updated 2026-09-29)
Quality tracking (Livenerf)Not publishedWindow 1 (days 11-20) score 0.592 vs baseline 0.584, delta +0.007 plus or minus 0.018: no change
Terminal-Bench 4.0 (official board)53.94%at xhigh (rank 11), 51.82% at max, Claude Code, 330 trials64.85% (plus or minus 3.11)rank 1, Claude Code, max effort, 330 trials, $4,718.65 total
Terminal-Bench 4.0 (Vals AI, same harness for all)53.53%rank 8, $18.60 per test65.15%rank 1, $13.20 per test
Arena textRank 4 (1508)Opus 5 (Max), 34,526 votesRank 2 (1516)Opus 5.5 (High), 10,322 votes
Arena WebDevRank 8 (1690)Opus 5 (Max), 18,924 votesRank 1 (1807)Opus 5.5 (Max), 3,119 votes
Epoch Capabilities Index163.94(CI 161.12 to 168.24)168.83(95% CI 165.25 to 173.95), highest in the file

Scores appear side by side only when both were published for the same test. Settings differ between sources, so small gaps can reverse under another harness.

Updates

Both Claude Opus 5 and Claude Opus 5.5

Claude Opus 5

5 recorded incidents in the last 90 days.

Retiring

Copilot to remove Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra on November 16

Copilot will deprecate Claude Opus 4.8 and Claude Opus 5 (replace with Claude Opus 5.5) and GPT-5.6 Terra (replace with GPT-6.1 Sol or GPT-6 Luna) on November 16, 2026. Opus 4.8 stays for individual annual-plan subscribers.

Takes effect Nov 16, 2026 · in 15 days

Who is affected:
Copilot users and integrations pinned to those models.
Action needed:
Move to the suggested alternatives before November 16, 2026.

Benchmark

Vals AI retires SWE-bench Verified with Opus 5 first at 97.0%

Vals AI archived its SWE-bench Verified leaderboard because performance has saturated. As of September 1, 2026, Claude Opus 5 with Mini-SWE-agent ranks first at 97.00%, ahead of DeepSeek V4 Pro 0813 (96.40%) and GPT-5.6 Sol (96.20%). Newer models such as Opus 5.5 and GPT-6 are not run.

Who is affected:
Anyone citing SWE-bench Verified to compare current models.
What to do:
Use Terminal-Bench 4.0 or SWE-Bench Pro results for newer models.

Claude Opus 5.5

3 recorded incidents in the last 90 days.

Now available

Full Claude Platform available on Google Cloud (Anthropic managed)

Claude Models on Google Cloud (Anthropic managed) is generally available with US and Global endpoints. It gives Google Cloud customers the full Claude Platform feature set with Google Cloud IAM, billing and commitment drawdown. Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 are available, and fast mode for Opus 5.5 costs $8 / $40 per million input and output tokens there.

Who is affected:
Teams that buy Claude through Google Cloud and want features that the Google-managed offering lacks.
What to do:
No action needed. Google Cloud customers who need same-day Claude API features can switch to the Anthropic-managed offering.

Switching checklist

  1. Change the model ID from claude-opus-5 to claude-opus-5-5 (on Bedrock, anthropic.claude-opus-5-5).
  2. Remove thinking set to disabled or enabled. Thinking cannot be turned off on Opus 5.5, so choose an effort level instead.
  3. Set effort explicitly: Opus 5.5 defaults to medium, where Opus 5 defaults to high.
  4. Replace tool_choice any or tool with auto plus strict tool use or structured outputs. Forced tool use returns an error on Opus 5.5.
  5. Read response content blocks by type, and pass thinking blocks back unmodified in tool-use loops.
  6. If your interface shows the short notes between tool calls, set thinking.display: they now arrive as progress-update thinking blocks, not text.
  7. In Claude Code, the /claude-api migrate skill can make these code changes for you.
  8. Rerun your evaluation set and compare cost per finished task, not cost per token.

FAQ

Is Claude Opus 5.5 cheaper than Opus 5?

Yes. Opus 5.5 costs $4 input and $20 output per million tokens, and Opus 5 costs $5 and $25, so Opus 5.5 is 20% cheaper at list price.

Is Opus 5.5 better at coding than Opus 5?

On Terminal-Bench 4.0 with Claude Code as the agent, Opus 5.5 scored 64.85% and Opus 5 scored 53.94%. Run through the API in Vals AI's Mini-SWE-agent harness, they scored 65.15% and 53.53%.

When will Claude Opus 5 be retired?

No retirement is announced. Anthropic lists Opus 5 as legacy, retiring no sooner than July 24, 2027, and gives at least 60 days' notice.

Can I turn off thinking on Opus 5.5?

No. A request with thinking disabled returns a 400 error on Opus 5.5. Opus 5 still allows it.

Has Opus 5.5 gotten worse since launch?

Not by Livenerf's measure. It tests Opus 5.5 through Claude Code every day and found no change in its first 10-day comparison window (+0.007, plus or minus 0.018). Its test can only detect drops of about 6.6 points or more.

Do I need to change my code to use Opus 5.5?

Usually yes. Besides the model ID, Opus 5.5 rejects disabled thinking and forced tool_choice, and defaults to medium effort where Opus 5 defaults to high.

Which has the newer knowledge cutoff?

Opus 5.5: June 2026, against May 2026 for Opus 5.

Related comparisons

Claude Opus 5.5 vs GPT-6 AstraClaude Opus 4.8 vs Claude Opus 5Claude Sonnet 5.5 vs Claude Opus 5.5Claude Opus 5 pageClaude Opus 5.5 page

Sources and method

Prices, specs and limits come from each maker's own pages. Benchmark scores come from independent benchmark publishers and the makers' announcements, with the setting each was measured in. Changes come from changelogs, deprecation pages and status pages. The summary is written from these facts only.