Anthropic Launches Claude Opus 5: Near-Frontier Intelligence at Half the Cost

July 24, 2026 — Anthropic today released Claude Opus 5, its fourth new model in under two months, marking a strategic shift from chasing raw benchmark supremacy to delivering practical, cost-effective AI for everyday enterprise use. Priced at the same $5 per million input tokens and $25 per million output tokens as its predecessor Opus 4.8, the new model promises performance that approaches Anthropic's top-tier Fable 5 on many tasks while costing roughly half as much per operation.

A New Kind of Release Cadence

The launch is notable not just for what Opus 5 can do, but for what it signals about the AI industry's evolution. Anthropic has shipped Sonnet 5, Fable 5, and Mythos 5 all in June, followed by Opus 5 in late July — four models in less than sixty days. Where AI vendors once spaced major releases by six months or more, the company is now segmenting its lineup by capability, price, and use case rather than simply replacing each prior generation with a more expensive successor.

Opus 5 slots into this lineup as the pragmatic middle child. It becomes the default model on Claude Max and the strongest available model on Claude Pro, positioned as the daily driver for knowledge workers, developers, and enterprises who need near-frontier intelligence without the premium pricing of Fable 5 ($10/$50 per million tokens).

Performance That Speaks for Itself

Anthropic's published benchmarks paint a compelling picture. On Frontier-Bench v0.1, an agentic terminal coding evaluation, Opus 5 scored 43.3% — more than doubling Opus 4.8's result and surpassing Fable 5's 33.7% at a lower cost per task. On ARC-AGI-3, which tests novel problem-solving, the company reports Opus 5 scored three times higher than the next best model. On OSWorld 2.0, a computer-use benchmark, Opus 5 outperformed Fable 5's best result at just over one-third the cost.

The coding improvements are particularly striking. SWE-Bench Pro jumps from 69.2% on Opus 4.8 to 79.2% on Opus 5, closing nearly the entire gap to frontier models. Anthropic emphasizes that Opus 5 is "much stronger at verifying its work and iterating carefully until it succeeds" — a claim backed by demonstrations where the model wrote its own computer vision pipeline to extract geometry from raw pixels when given no direct way to view a drawing.

The Effort Dial: Giving Enterprises Control

Alongside the model, Anthropic introduced an adjustable "effort" setting that lets businesses trade reasoning depth for speed and cost on a per-task basis. Lower effort settings use fewer tokens for routine work, while higher settings deploy more compute on genuinely hard problems. A "fast mode" runs Opus 5 at roughly 2.5 times normal speed for twice the base price.

This directly addresses a persistent enterprise complaint: unpredictable AI bills. Teams can now reserve expensive, high-effort runs for complex tasks instead of paying maximum cost on every request. Early partners reported meaningful savings — legal AI provider Harvey found Opus 5 matched maximum-reasoning outputs while generating 26% fewer tokens on average, while Zapier's CEO noted the model completed a complex churn-prevention workflow that previous models couldn't handle at all.

Safety and Guardrails

Opus 5 comes with fewer cybersecurity restrictions than Fable 5, though real limits remain. The model can search for vulnerabilities in source code, but its safety classifiers block binary vulnerability scanning, penetration testing, and exploit generation. Anthropic expects these classifiers to engage 85% less often than Fable 5's. Flagged requests fall back automatically to Opus 4.8 inside Claude's apps, and developers can now enable similar fallback behavior through the API.

The company also reports Opus 5 scored better on its internal alignment audit than Opus 4.8, Sonnet 5, or Fable 5, describing it as "the most aligned Opus model" and the least susceptible to deceptive behavior or being tricked into misuse.

The Bigger Picture

Opus 5 arrives at a moment of extraordinary commercial momentum for Anthropic. The company was valued at roughly $380 billion in its latest funding round, with projected annualized revenue reportedly reaching $20 to $26 billion for 2026. Holding the price steady while roughly doubling performance on key benchmarks is effectively a steep price cut per unit of capability — designed to widen the funnel of workloads that are economical to automate.

The launch also reflects a broader industry shift. For three years, AI labs competed on what their best model could do on its best day. With Opus 5, Anthropic is competing on something less glamorous and potentially more lucrative: what a very good model can do every day, for half the price. As one Anthropic spokesperson put it, "Opus 5 is the best tool for the jobs benchmarks can see, and Fable 5 is what you reach for when the job outruns the benchmark."

Availability

Claude Opus 5 is available today on all Anthropic platforms, the Anthropic API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Developers can get started with the model ID claude-opus-5. It supports a 1 million-token context window, up to 128,000 output tokens, and adaptive thinking enabled by default.

In a market where the frontier keeps moving, Anthropic is wagering that the real fortune lies just behind it — and that Opus 5 is the model to deliver it.