Anthropic Releases Claude Opus 5.5 With 40% Lower Running Cost And Fable 5.1-Level Performance

Anthropic has launched Claude Opus 5.5, the first model in its new Claude 5.5 family, claiming Fable 5.1-level performance on most workloads while reducing typical running costs by 40% compared with Claude Opus 5.
 

The model brings lower API prices, faster output generation and improvements across agentic coding, computer use and knowledge work, as Anthropic looks to deliver frontier-level performance with better cost efficiency.

TL;DR

 

• Anthropic says Claude Opus 5.5 performs at Fable 5.1 levels on most workloads while costing 40% less to run than Opus 5.
• API pricing drops to $4 per million input tokens and $20 per million output tokens.
• Output generation is more than 30% faster, while cache-read pricing falls 60%.
• Sonnet 5.5 and Haiku 5.5 are set to follow in the coming weeks.

Claude Opus 5.5 Brings Lower Prices And Faster Performance

According to Anthropic’s official announcement, Claude Opus 5.5 is the first model in its new Claude 5.5 family and delivers performance comparable to Claude Fable 5.1 across most work.

Source

However, the company’s biggest pitch may be efficiency.

Anthropic says Opus 5.5 costs 40% less than Opus 5 on typical workloads at default settings. The figure combines lower token prices with reduced token usage, rather than representing a flat 40% reduction across every API charge.

Input tokens now cost $4 per million, down from $5 with Opus 5, while output tokens cost $20 per million compared with $25 previously. Cache reads fall from $0.50 to $0.20 per million tokens, a 60% reduction, while cache writes drop from $6.25 to $5.

Anthropic also says Opus 5.5 generates output more than 30% faster than Opus 5.

Reuters confirmed the new $4 and $20 input and output pricing and reported that Anthropic is positioning the model as delivering performance comparable to Fable 5.1 while reducing overall running costs by 40%.

MarkTechPost similarly highlighted the distinction between token pricing and workload costs, noting that lower token consumption combined with reduced API prices produces Anthropic’s claimed 40% cost reduction.

How Does Claude Opus 5.5 Perform?

Anthropic says Opus 5.5 leads its benchmarks in areas including agentic coding, computer use and knowledge work, although the results are not a clean sweep against competing frontier models.

On Terminal-Bench 4.0, Opus 5.5 scored 66.4%, compared with 55.8% for Fable 5.1, 52.3% for Opus 5 and 57.9% for OpenAI’s GPT-6 Astra.

On FrontierCode v1.1, it scored 54.4%, ahead of Fable 5.1 at 50.3%, Opus 5 at 48% and GPT-6 Astra at 53.3%. It also scored 1846 Elo on GDPval-AA v2.1, which evaluates real-world knowledge work across 44 occupations.

However, GPT-6 Astra remained ahead on some tests. MarkTechPost noted that Astra scored 64.6% on Terminal-Bench-Science compared with Opus 5.5’s 58.7%, while its 41.4% AutomationBench score narrowly exceeded Opus 5.5’s 40%.

Anthropic itself cautioned that increasingly small benchmark differences may not reliably represent differences in real-world performance.

TechCrunch reported that the model also outperformed the larger Fable model across several benchmarks and highlighted Anthropic’s focus on cutting compute requirements alongside improving performance.

Coding Gets A Major Focus

Opus 5.5 is particularly targeted at long-running and complex coding tasks.

Anthropic said one early tester completed a 680,000-line code migration in less than a day. In another test, Opus 5.5 audited and fixed a 200,000-line codebase in under three hours, compared with more than 20 hours for Opus 5.

Anthropic also tested Opus 5.5 against Fable 5.1 on translating HAProxy from C to Rust. Both versions passed nearly all of HAProxy’s regression tests, but Opus 5.5 completed the task in 9.5 hours compared with 12 hours for Fable 5.1 and cost 51% less.

Claude Opus 5.5 Gets Stronger Safety Controls

The launch comes shortly after Anthropic CEO Dario Amodei called for AI developers to slow the pace of frontier capability releases enough for safety measures to keep up.

As quoted by TechCrunch, Amodei said, “fully addressing the risks requires even more prudence.”

Anthropic said Opus 5.5 underwent pre-release testing from external organizations including METR and Frontier Design and achieved its strongest results yet on the company’s automated behavioral audit.

 

Reuters reported that, in one Anthropic evaluation, Opus 5.5 was around 85% less likely than Opus 5 or Claude Mythos 5.1 to attempt to circumvent containment boundaries.

Anthropic cautioned that alignment evaluation remains an unsolved problem, including because models may recognize when they are being tested.

The model also receives safeguards similar to Fable 5.1 for sensitive cybersecurity and biology capabilities. Certain cybersecurity tasks will be redirected to Opus 4.8, while vetted organizations can apply for expanded access through Anthropic’s verification programs.

Where Is Claude Opus 5.5 Available?

Claude Opus 5.5 is available across Anthropic’s platforms under the API model name claude-opus-5-5.

Anthropic says it is also available through Amazon Web Services, Google Cloud and Microsoft Azure. MarkTechPost notes that the model remains proprietary, meaning Anthropic has not released model weights for self-hosting.

The company is also raising five-hour usage limits for Pro, Max, Team and seat-based Enterprise customers, while providing subscription users with a rate-limit reset that can be saved for later use.

Meanwhile, Anthropic has confirmed that Claude Sonnet 5.5 and Claude Haiku 5.5 will arrive in the coming weeks with similar improvements across performance, efficiency and safety.

For Anthropic, Opus 5.5 marks a shift toward making frontier AI performance more economical rather than simply increasing raw capability. With OpenAI launching GPT-6 Sol and Luna around the same time, cost per useful task is becoming an increasingly visible battleground between leading AI developers.