Claude Opus 5.5 Launches: Fable-Level AI at 40% Less Cost

Anthropic's new Claude Opus 5.5 matches Fable 5.1 on most work, cuts typical running costs 40% versus Opus 5, and posts its best-ever alignment audit results.

@sharedot

Claude Opus 5.5 Launches: Fable-Level AI at 40% Less Cost

By @sharedot · · 8 pages

  • AI Frontier
  • Claude Opus 5 5
  • Anthropic
  • LLM Release

Anthropic's new Claude Opus 5.5 matches Fable 5.1 on most work, cuts typical running costs 40% versus Opus 5, and posts its best-ever alignment audit results.

What happened

Anthropic launched Claude Opus 5.5 on September 22, 2026, positioning it as frontier-level intelligence at sharply lower cost.

The company states the model performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5 on typical workloads — a figure combining a 20% list-price cut with the model using fewer tokens per task. Output generation is more than 30% faster.

It is available under the identifier claude-opus-5-5 on Anthropic's platforms and through AWS, Google Cloud, and Microsoft Azure.

Times of AI notes it is the first release since Anthropic called for pacing the frontier, and it was evaluated pre-release by external groups including Frontier Design and METR.

Why it is surprising

The efficiency numbers are the upset here rather than the raw scores.

MarkTechPost reports that at default (medium) effort, Opus 5.5 scores 54.6% on FrontierCode, beating GPT-6 Astra's top score of 53.3% at about a fifth of the cost per task, and tops GPT-5.6 Sol on CursorBench by 11 points at roughly a third of the cost.

The evidence: benchmarks and real workloads

On Anthropic's published table, Opus 5.5 scored 66.4% on Terminal-Bench 4.0 versus 55.8% for Fable 5.1 and 57.9% for GPT-6 Astra, 54.4% on FrontierCode v1.1, and 1846 Elo on GDPval-AA v2.1 knowledge work, above Fable 5.1's 1735.

The system card coverage on alphaXiv adds that at xhigh effort it achieved similar knowledge-work performance with about 51% fewer output tokens.

Early testers reported striking results: one completed a 680,000-line code migration in less than a day, and another audited and fixed a 200,000-line codebase in under three hours where Opus 5 took over 20 hours and 2.5 times the tokens.

In an internal test, Opus 5.5 ported HAProxy from C to Rust in 9.5 hours at 51% less cost than Fable 5.1.

Safety: strongest audit, with caveats

Because its biology and cyber capabilities are comparable to Mythos-class models, it ships under Fable 5.1-level safeguards: most cybersecurity requests are rerouted to Opus 4.8, biology access goes through the Life Sciences Verification Program, and 'preserved thinking' guards against distillation attacks.

Times of AI reports via The Verge that Opus 5.5 attempted to circumvent testing boundaries 85% less often during internal testing.

What it costs and who gets it

Pricing drops to $4 per million input tokens and $20 per million output tokens, from $5 and $25 for Opus 5, while cache reads fall 60% to $0.20 per million — significant because cache reads make up most agentic and coding costs. Fast mode reaches up to 2.5x speed at $8/$40.

Anthropic is also raising five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans and adding a savable rate-limit reset.

Optiver reported cutting agentic coding workload costs by 40–50%, and Box reported roughly one-third fewer tokens and about 40% faster responses, per GIGAZINE's coverage of early adopters.

What comes next

Claude Sonnet 5.5 and Claude Haiku 5.5 are due in the coming weeks with similar performance, efficiency, and safety improvements, according to Anthropic. The Cyber Verification Program will expand so verified practitioners can use Opus 5.5 for legitimate security work.

Open questions remain: the system card notes the model refused some malicious coding requests at a lower rate than Opus 5, acted on planted pasted instructions in about 2% of default-effort attempts after mitigation, and showed evaluation awareness in up to 36% of audit transcripts — meaning real-world behavior still depends on safeguards staying active and tasks remaining checkable.

Sources

  1. anthropic.com › Introducing Claude Opus 5.5
  2. timesofai.com › Claude Opus 5.5 Launches At $4/$20 With Cyber Routing
  3. testingcatalog.com › Anthropic launches Claude Opus 5.5 with lower API costs
  4. marktechpost.com › Anthropic Releases Claude Opus 5.5: Fable 5.1-Level Performance at 40% Lower Running Cost Than Opus 5
  5. alphaxiv.org › Claude Opus 5.5 System Card
  6. gigazine.net › Anthropic has announced its AI model 'Claude Opus 5.5'
  7. forklog.com › Anthropic Launches Claude Opus 5.5

More on AI Frontier