Claude Opus 5 Launches With Near-Frontier Intelligence at Half the Price, Landing in GitHub Copilot on Day One

Claude Opus 5 Brings Near-Frontier Intelligence at Half the Cost, Landing in GitHub Copilot on Day One

Anthropic's newest model, Claude Opus 5, arrived today in two places at once: at the top of its product lineup and inside the code editor millions of developers already use. The launch's sharpest edge is the economics. Opus 5 matches the benchmark performance of Anthropic's flagship Claude Fable 5 at roughly half the cost per task 1. It shipped directly into GitHub Copilot across every major development environment on launch day, putting near-frontier AI into existing developer workflows 2.

The benchmark numbers are Anthropic's own, measured on evaluations the company designed or adopted. On Frontier-Bench v0.1, Opus 5 more than doubles the performance of its predecessor, Opus 4.8, at a lower cost per task, and surpasses every other model in Anthropic's testing 1. On CursorBench 3.2 at maximum effort, Opus 5 lands within half a percentage point of Fable 5's peak score while costing half as much per task 1.

Opus 5 is the lower-cost model that gets close enough to matter. It costs half as much to get within striking distance 1.

Third-party evaluations were not available at launch. Every benchmark figure here carries the caveat that Anthropic both ran the tests and reported the results. That is standard practice for new model releases, but it matters because the price-performance ratio is the entire pitch. Independent eval runs will determine whether the gap to Fable 5 is as narrow as Anthropic claims.

The distribution move is harder to copy. GitHub added Opus 5 to Copilot across Visual Studio Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, the Copilot cloud agent, GitHub Mobile, and github.com, all on launch day 2. GitHub described the model as suited for complex, long-running coding tasks requiring careful reasoning and noted particular strength in autonomous code changes, regression verification, and multi-tool coordination 2. The model is available to Copilot Pro+, Max, Business, and Enterprise subscribers, with gradual rollout underway 2. Copilot Enterprise and Business administrators must enable the model in settings before their teams can access it 2.

Developers who already pay for Copilot can switch to Opus 5 from a model picker. No migration, no new billing relationship, no separate API integration. The model is billed at provider API list price under GitHub's usage-based billing 2.

Anthropic emphasizes that Opus 5 does something its predecessors struggled with: verifying its own work. In early-access testing reported by Anthropic, the model independently validated outputs, debugged root causes rather than surface symptoms, and built its own test harnesses when none existed 1. In one example, when handed a drawing of a machine part with no way to view the image directly, Opus 5 wrote its own computer vision pipeline to extract geometry from raw pixels and reconstructed the part in 3D 1. In another, it found the root cause of a bug in a popular open-source package manager that a competing model missed, fixing an edge case the community's own patch had overlooked 1.

These are Anthropic's own examples and customer testimonials, not independent verification. But they describe a behavioral pattern: a model that loops back to validate its work before declaring success, rather than producing a first draft and moving on. GitHub's early testing echoed the observation, citing the model's effectiveness at validating its work and reducing unnecessary execution overhead on complex tasks 2.

Opus 5 ships with cybersecurity guardrails that could affect developer workflows. Anthropic itself notes that the model trails another model, Mythos 5, on cybersecurity tasks 1. GitHub separately warns that the model includes "enhanced safeguards for high-harm cyber content" that "may block some cyber-related or security-adjacent requests" 2. GitHub's guidance for blocked requests is to rephrase with additional benign context or select another available model 2.

For developers whose work touches security analysis, vulnerability research, or incident response, that filter could interrupt legitimate workflows. Neither Anthropic's announcement nor GitHub's changelog specifies how the safeguards distinguish offensive from defensive cybersecurity work.

Opus 5 is the default model on Claude Max and the strongest model on Claude Pro 1.

For builders, the arithmetic is plain. If Anthropic's benchmarks hold up under independent testing, a model that performs within half a percentage point of the flagship on coding tasks, at half the cost, is available inside the IDE you already have open. The economics of AI-assisted development shifted today. The question is whether competitors can match it before the gap widens.