Key facts
- Anthropic released its Claude Opus 5.5 AI model on Tuesday.
- Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens.
- The new model is 40% cheaper than Opus 5 at default settings.
- Anthropic claims Opus 5.5 matches Fable 5.1 on most tasks.
- GPT-6 Astra still outperforms Opus 5.5 on AutomationBench and Terminal-Bench-Science 0.1.
- Opus 5.5 carries the same cybersecurity and biology safeguards as Fable 5.1.
Anthropic launched its Claude Opus 5.5 AI model on Tuesday, positioning it as a significant advancement in AI capabilities while also reducing costs for users. The company claims the new model matches the performance of its top-tier Fable 5.1 model on most tasks, despite being priced more competitively.
Opus 5.5 is available at $4 per million input tokens and $20 per million output tokens, representing a 40% price reduction compared to the previous Opus 5 model. Cache reads, a significant cost driver for long agentic coding sessions, have also been reduced by 60% to $0.20 per million tokens.
While Opus 5.5 leads Fable 5.1 and Opus 5 on benchmarks for coding and knowledge work, Anthropic acknowledges that GPT-6 Astra still holds an edge on specific tests like AutomationBench and Terminal-Bench-Science 0.1. The model's performance on Terminal-Bench 4.0 was 66.4%, surpassing Fable 5.1's 55.8% and GPT-6 Astra's 57.9%. On the GDPval-AA v2.1 score, Opus 5.5 achieved 1846, compared to Fable 5.1's 1735 and Opus 5's 1708.
This release is the first from Anthropic since CEO Dario Amodei publicly advocated for a slowdown in AI capability advancements to allow safety and alignment measures to keep pace. Opus 5.5 incorporates the same cybersecurity and biology safeguards as Fable 5.1, with most cybersecurity tasks being rerouted to the older Opus 4.8 model. Anthropic plans to release Sonnet 5.5 and Haiku 5.5 in the coming weeks, which are expected to feature similar price cuts.