Anthropic launches Claude Opus 5.5, matching Fable 5.1 performance at ~40% lower operating cost with faster output and less "Claudish" writing

Anthropic released Claude Opus 5.5 (model ID claude-opus-5-5), available now on the Claude Platform, AWS, Google Cloud and Microsoft Azure. It cuts token prices 20% to $4/$20 per million input/output tokens (from $5/$25 for Opus 5), cuts cache-read costs 60%, raises five-hour usage limits 20% for subscribers, and Anthropic claims it matches Fable 5.1 benchmark performance while using fewer tokens and generating output over 30% faster. The release also adds safety routing for cybersecurity, biology and frontier-LLM queries to other models, introduces anti-distillation "Preserved Thinking" measures and EU AI Act watermarking, and removes the option to disable Thinking mode.

This is directly actionable for anyone budgeting API spend or picking a model: lower per-token pricing plus reduced token usage compounds into a real cost cut for high-volume users, and the higher usage caps matter for subscribers hitting rate limits. Enterprises and developers building on Opus should re-run their own cost and latency comparisons rather than accept Anthropic's figures, since the pricing pressure also signals a broader industry trend of frontier labs competing on cost as much as capability. The removal of the "disable Thinking" toggle is a real workflow change for anyone who relied on non-reasoning mode for speed or simpler outputs, and this is not being flagged prominently in coverage.

The "40% lower operating cost" figure and the "less Claudish" writing claim are self-reported and subjective, not independently verified in the sourcing available here. The headline framing of "matching Fable 5.1 performance" implies parity, but the underlying benchmark comparison, its margin of error, and which specific tasks were tested are not detailed in this evidence, so treat it as a vendor claim pending third-party confirmation.

1. 01Independent benchmark confirmation: check whether Artificial Analysis or another third party verifies the claimed ~40% lower operating cost and 30%+ faster output, not just Anthropic's own numbers
2. 02OpenAI and Chinese lab pricing response over the next 4-8 weeks, since Opus 5.5's $4/$20 pricing only matters if it holds up against GPT-6 Astra/GPT-5.6 Sol and cheaper alternatives
3. 03Developer reaction to the removal of the 'disable Thinking' option and the new safety-routing for cybersecurity/biology/frontier queries, which could affect latency, cost or output for specific workloads not captured in headline benchmarks