What did Anthropic do?! (Opus 5)

2026-07-24 · Source: Matthew Berman · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, long

Summary

Anthropic has released Claude Opus 5, a new large language model that significantly outperforms its predecessor, Fable 5, on numerous benchmarks while costing half the price. Opus 5 achieved scores like 43 on Aentic terminal coding, a 100-point improvement on GDP val, and a 30% success rate on Arc AGI 3, compared to Fable 5's 33, lower score, and 8% respectively. It also demonstrates superior efficiency, offering higher performance at a lower cost per task than Fable 5 and competitive pricing with GPT 5.6 Soul. Priced at \$5 per million input tokens and \$25 per million output tokens, Opus 5 shows substantial gains in complex enterprise knowledge work, such as a 6-point jump in data analysis on the Box AI benchmark. While stronger than Opus 4.8 in cybersecurity, it remains behind Mythos 5, likely due to integrated guardrails. Opus 5 is available now, featuring automatic fallbacks for safety-flagged requests.

Key takeaway

For Machine Learning Engineers evaluating new LLMs for deployment, Claude Opus 5 presents a compelling option. You should prioritize evaluating models based on "cost per task" rather than just token price, as Opus 5 offers significantly better performance and efficiency than Fable 5 at half the cost. Consider integrating Opus 5 for complex analytical tasks and potentially pairing it with Fable for advanced planning, but be aware of potential safety filter fallbacks.

Key insights

Anthropic's Claude Opus 5 delivers superior performance and cost-efficiency compared to Fable 5, setting a new benchmark for LLMs.

Principles

• Cost per task is the critical metric for LLM evaluation.
• Guardrails can improve overall model performance despite capability reduction.
• Model optimization can significantly enhance efficiency post-initial release.

In practice

• Evaluate LLMs using cost per task, not just token price.
• Consider Opus 5 for multi-step analytical enterprise tasks.
• Pair Opus 5 with Fable for complex planning or debugging.

Topics

• Claude Opus 5
• LLM Benchmarking
• Cost Per Task
• Enterprise AI
• Model Efficiency
• AI Safety Guardrails

Best for: CTO, VP of Engineering/Data, AI Engineer, AI Scientist, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Anthropic Shipped Four Models in Nine Weeks. Opus 5 Explains the Other Three. — Claude Opus 5 offers near-frontier intelligence at half the price of Fable 5, impacting the value proposition of other models.
• Anthropic's Opus 5 is about token efficiency, not a capability leap — Anthropic's Opus 5 prioritizes cost-efficiency and iterative performance gains over a major capability breakthrough.
• Anthropic's Claude Opus 5 costs well below Fable 5 while matching or beating it across most benchmarks — Claude Opus 5 leads benchmarks in capability and cost-efficiency, but its high hallucination rate demands careful application.
• A Cheaper Claude, a Cracked Agent, and $52 Million for Your Iris — Frontier AI models are achieving higher performance at significantly reduced costs, making advanced capabilities more accessible.
• Anthropic's Opus 5 surprise — Claude Opus 5 delivers Fable 5-tier intelligence at half the cost, setting new benchmarks in agentic tasks and complex problem-solving.
• OPUS 5 CLICK NOW — Claude Opus 5 redefines LLM value, offering superior performance at a significantly lower cost per task.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Matthew Berman.