Anthropic pushed Claude Opus 5 to all platforms on July 24. The core narrative in one sentence — the price is unchanged from Opus 4.8, but it runs within 0.5% of Fable 5's peak score on CursorBench 3.2, and surpasses Fable 5's best OSWorld 2.0 result at roughly one-third of the cost.
What's worth unpacking is how it solidifies the "gradient pricing" of Anthropic's July model lineup. Opus 5 costs $5 input / $25 output per million tokens; Fable 5, inferred from multiple third-party platforms, sits at the $10 / $50 tier. Opus 5 hit SOTA on all four benchmark cards — CursorBench 3.2 max effort, Frontier-Bench v0.1, Zapier AutomationBench, and OSWorld 2.0. On Frontier-Bench v0.1 it delivered more than twice Opus 4.8's performance at lower per-task cost. On ARC-AGI 3 it scored roughly three times the second-place model.
More notable are the numbers outside coding: +22% on internal agentic coding evals versus Opus 4.7; financial modeling accuracy up 9 percentage points with one-third fewer turns and 60% less time; legal agent tasks consuming 26% fewer tokens than Opus 4.8 max reasoning. These aren't vague "better than Opus 4.8" claims — they're differences you can convert into tool-call counts and wall-clock time.
But a few boundaries need to be spelled out:
1. Opus 5's cybersecurity classifier intervenes 85% less than Fable 5's, but its exploitation capability is still far behind Mythos 5 — it's only allowed to find vulnerabilities at the source-code level; binary scanning and penetration testing remain blocked by default. 2. Its overall misalignment behavior score is 2.3, the lowest of recent generations; this widens the window of tasks you can safely delegate to longer chains, but the official context window figure wasn't given. 3. The +22% agentic coding figure over Opus 4.7 comes from internal evals whose specifics weren't published; for SWE-bench-style public comparisons, we'll need to wait for external leaderboards.
Companion feature updates worth noting: the Claude API now offers "mid-conversation tool switching" (without invalidating the prompt cache) and "automatic fallback" (auto-routing to other models when blocked by a classifier). Both are details that reduce incident rates for people running agents in production.
My own take: Opus 5 is the "underrated one" of Anthropic's July lineup. Fable 5's halo is too bright, and Opus 4.8 had attention stolen by Mythos 5. But any engineering team that has run CursorBench / OSWorld / Frontier-Bench will rethink their budget once they see this pricing curve. Opus 5 is a key rung in the "model gradient convergence" of 2026 H2 — for the first time, Anthropic simultaneously has Fable 5 / Mythos 5's ceiling and Opus 5's affordable floor.