Claude Opus 5 Brings Frontier Coding at Half the Cost

Anthropic released Claude Opus 5 on July 24, 2026, and the number that jumps out is the price. The new model delivers near-frontier coding and reasoning at half the cost of the company's flagship Fable 5, landing at $5 per million input tokens and $25 per million output tokens. It is available immediately across Claude.ai, the API, Claude Code, and Claude Cowork, and it takes over as the default model on the Claude Max tier.

• Priced to scale: $5 per million input tokens and $25 per million output — roughly half Fable 5's cost for comparable quality
• Big context: a 1M-token context window plus a new "xhigh" reasoning effort mode for the hardest problems
• Available now: live on Claude.ai, the API, Claude Code, and Claude Cowork; the default on Claude Max and the strongest model on Claude Pro
• Benchmark gains: Anthropic reports Opus 5 more than doubles Opus 4.8 on Frontier-Bench v0.1 and comes within 0.5% of Fable 5 on CursorBench 3.2

What Makes Claude Opus 5 Different?

Opus 5 is the fourth model in the Claude 5 family, joining Sonnet 5, Fable 5, and Mythos 5. Where Fable 5 targets days-long autonomous projects at the very frontier, Opus 5 is tuned for the everyday heart of the market: coding, enterprise automation, and agentic workflows that need to run reliably and affordably at scale. Anthropic frames it as capturing nearly all of Fable 5's intelligence at half the price — a signal that the industry's focus is shifting from peak capability on a single hard problem toward efficiency and cost-per-task across many. For teams building on Claude, that shift translates directly into more useful work per dollar.

Where Does Claude Opus 5 Shine?

Coding is the clearest win. Anthropic reports that Opus 5 more than doubles the previous Opus 4.8 on its Frontier-Bench v0.1 software-engineering suite, and lands within 0.5% of Fable 5 on the CursorBench 3.2 evaluation while costing half as much. On the AA Coding Agent Index, the company describes state-of-the-art results. The 1M-token context window is a natural fit for long-running agents that need to hold an entire codebase or a lengthy task history in working memory, and the optional "xhigh" reasoning mode lets developers dial up deliberation for the trickiest problems.

Strong Results in Science and Knowledge Work

The gains extend well beyond code. Anthropic says Opus 5 scores roughly three times higher than the next-best model on the ARC-AGI 3 reasoning benchmark, and posts double-digit improvements on scientific tasks — up 10.2 percentage points over Opus 4.8 on organic chemistry and 7.7 points on protein-sequencing work. The company also calls Opus 5 its "most aligned model to date" based on internal behavioral audits, a reassuring note for enterprises moving agents into production.

Why the Price Cut Matters

Halving the cost of near-frontier intelligence is the real headline. Cheaper high-quality inference lowers the barrier for startups, researchers, and small teams to build ambitious agentic products — the kind of work that used to sit behind flagship pricing. It also dovetails with Anthropic's broader compute push; see our coverage of AMD's multi-billion-dollar investment in Anthropic's AI infrastructure and the wider wave of accessible AI tooling such as OpenAI's new ChatGPT program for small businesses. For more on model releases and what they mean for builders, follow our ongoing AI coverage.

The takeaway is simple: Opus 5 makes the most capable tier of Claude dramatically more affordable, and that is exactly the kind of progress that puts genuinely powerful tools into more hands.

Sources: Anthropic — July 24, 2026; Bloomberg — July 24, 2026; Yahoo Finance — July 24, 2026.