Anthropic launches Claude Opus 5 at half Fable 5's cost

Anthropic released Claude Opus 5 on July 24-25, 2026, positioning the model as a cost-effective alternative to its flagship Fable 5. Priced at $5 per million input tokens and $25 per million output tokens—half the cost of Fable 5—Opus 5 delivers comparable or superior performance on numerous benchmarks, particularly in software engineering, knowledge work, and autonomous reasoning tasks.

On coding-focused evaluations including Frontier-Bench and GDPval-AA, Opus 5 ranks among the top performers, surpassing Fable 5 on several measures. On CursorBench 3.2 at maximum effort setting, it scores within 0.5% of Fable 5 while costing half as much per task. The model excels at iterative problem-solving and self-verification—it proactively checks work, detects errors, and re-attempts tasks until completion. Notable examples include writing a computer vision pipeline to reconstruct a 3D machine part from a pixel-level image with no direct viewing capability, and discovering root causes in open-source software bugs that community patches had missed.

Opus 5 introduces a configurable "effort" setting allowing users to optimize for reasoning strength or cost efficiency. It also features enhanced safety guardrails—automated behavior audits show it adheres more closely to Claude's usage guidelines than predecessors, with deceptive behavior reduced and resistance to manipulation increased. The model intentionally avoids specialized training on offensive cybersecurity tasks, remaining behind Mythos 5 (Anthropic's most capable restricted model) on exploit generation. Safety classifiers trigger approximately 85% less often for Opus 5 than Fable 5, and new beta features include dynamic tool modification mid-conversation and automatic fallback routing to lower-tier models when safeguards block requests.

The release reflects broader industry pressure on pricing. Anthropic reduced its system prompt for Claude Code by over 80% without measurable performance loss, suggesting improved model efficiency. The model became the default for Claude Pro subscribers and is available across all Claude platforms. Separately, Anthropic announced plans to deploy up to 2 gigawatts of AMD Instinct GPU systems by 2027, and AMD committed up to $5 billion in strategic equity investment. The release comes amid competition from lower-cost alternatives, including OpenAI's GPT-5.6 Sol and emerging open-weight models from China.