Anthropic launches Claude Opus 5, a cost-effective AI model for coding and enterprise workflows
The new AI model stands out in daily usage economics.
1Claude Opus 5 Launch
Anthropic announced Claude Opus 5, offering nearly all the intelligence of its top-tier Fable 5 model at half the cost. This signals a shift in the AI race from raw capability to daily usage economics. The model is immediately available across all Anthropic platforms.
2Pricing and Access
Opus 5 is priced at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8. It's positioned as the default model on Claude Max and the strongest option on Claude Pro. Anthropic emphasizes that economically critical AI work happens in mid-difficulty ranges.
3Performance Comparison
Anthropic claims Opus 5 sets new records on coding and knowledge-work benchmarks like Frontier-Bench and GDPval-AA. On Frontier-Bench v0.1, it scored 43.3%, more than double Opus 4.8's 18.7%. On ARC-AGI 3, it tripled the next best model's score.
4Limitations and Admissions
Opus 5 trails competitors in cybersecurity and biology research. Anthropic acknowledges its strength in bounded tasks but admits Fable 5 outperforms in long-duration work. This highlights what benchmarks don't measure: temporal coherence.
5Token Efficiency
Opus 5's key advantage is delivering high performance with fewer tokens. Harvey reported 26% fewer tokens than Opus 4.8. Fundamental Research Lab noted 60% less time spent on complex financial modeling tasks.
6Commercial Realities
Enterprise AI spending has moved beyond experimentation. Inference costs are now board-level concerns. For Anthropic's token-paying customers, doing more with fewer tokens isn't just nice—it's essential.
7Market Share and Revenue
Per Contrary Research's 2026 analysis, Claude held 40% of the enterprise LLM market by late 2025. Claude Code alone reached $1B annualized revenue, showcasing Anthropic's API and enterprise focus.
8Customer Experiences
Zapier CEO Wade Foster said Opus 5 topped their AutomationBench without extra tokens. Cognition's Scott Wu noted it delivers Fable-level performance on FrontierCode 1.1 at half the cost, excelling in debugging.
9Long-Term Autonomy
Anthropic positions Fable 5 for multi-day autonomous tasks requiring coherence across steps. Opus 5 excels in bounded tasks. This differentiation may define model competition in 2026 as benchmarks saturate.
10Self-Verifying Agents
Opus 5 can self-verify its work, aiming to reduce automation's hidden costs. This feature enhances reliability in complex workflows, minimizing human intervention needs.
11AI Race's New Phase
Opus 5's launch symbolizes the industry's shift from brute-force capability to efficiency. For enterprises, the smartest model matters less than the most economical one—a change that may reshape the sector permanently.