Anthropic Introduces Claude Opus 5 for Coding, Research, and Knowledge Work

What It Is

Claude Opus 5 is Anthropic’s new flagship model for everyday use—positioned as the default on Claude Max and the strongest model available on Claude Pro. It’s priced at $5 per million input tokens and $25 per million output tokens, matching its predecessor Opus 4.8 in cost while delivering substantially better results.

The pitch is straightforward: get close to Fable 5’s intelligence without paying Fable 5’s bill.

Where It Actually Performs

The benchmark story is specific enough to be useful here.

On Frontier-Bench v0.1, Opus 5 surpasses all other models and more than doubles Opus 4.8’s performance at the same cost. On CursorBench 3.2, it lands within 0.5% of Fable 5’s peak score at max effort—at half the cost per task.

Knowledge work results are similarly strong:

• ARC-AGI 3: Opus 5 scores three times higher than the next-best model on novel problem-solving
• Zapier AutomationBench: ~1.5× the pass rate of the next-best model at the same cost
• OSWorld 2.0: Outperforms every other model at any given cost, beating Fable 5’s best result at roughly a third of the price

For scientific research, improvements are most notable in organic chemistry and protein-related tasks—10+ percentage point gains over Opus 4.8 on internal benchmarks.

The Agentic Angle

Opus 5 isn’t just faster at answering questions—it’s noticeably better at working through problems without hand-holding.

A few examples from early-access testing stand out. Given a drawing of a machine part with no way to directly view it, Opus 5 wrote its own computer vision pipeline to extract geometry from raw pixels, then reconstructed the full 3D model. Competing models couldn’t solve the same task after five attempts.

On a real bug in a popular open-source package manager, Opus 5 found the root cause and fixed an edge case the community’s own patch had missed. A competing model fixed only the surface symptom and called it done.

This pattern—verify, iterate, don’t stop at the obvious answer—appears to be a deliberate design goal, not a side effect.

Safety and the Cybersecurity Tradeoff

Anthropic is transparent about where Opus 5 sits on the risk spectrum. It does not advance the frontier in dual-use capabilities. It remains behind Mythos 5 on offensive cybersecurity tasks—specifically on exploit development, even when vulnerability discovery is roughly comparable.

Cyber classifiers on Opus 5 are less restrictive than those on Fable 5, allowing source code vulnerability scanning while blocking binary-based scanning, penetration testing, and exploit generation. Flagged requests fall back to Opus 4.8 by default in Claude.ai, Claude Code, and Claude Cowork.

Enterprises already enrolled in Anthropic’s Cyber Verification Program get immediate access to a version with fewer restrictions. Related concerns around AI-assisted browsing and security are part of broader prompt injection risks.

Pricing and Practical Details

• Base price: $5 / 1M input tokens, $25 / 1M output tokens (same as Opus 4.8)
• Fast mode: ~2.5× default speed, available at 2× base price
• API model ID: claude-opus-5
• No data retention requirements for general access

Two beta features ship alongside the launch:

• Mid-conversation tool changes: Swap tools within a conversation without invalidating the prompt cache
• Automatic API fallbacks: Flagged requests route to the best available model instead of being blocked outright

The cost angle is central to the launch framing.

Who This Is For

Opus 5 is a strong fit for developers and teams doing serious coding, research, or multi-step automation work who’ve been priced out of Fable 5 or found Opus 4.8 hitting its ceiling. The cost-per-task framing matters here—if your workloads are high-volume, the efficiency gains compound quickly.

If you’re already on Claude Pro or Claude Max, the upgrade is automatic. If you’re on the API, it’s worth running your existing benchmarks against claude-opus-5 before assuming Fable 5 is still the right call for your use case. The gap may be smaller than you think—and the price difference is not. For teams comparing coding-focused tooling, Claude Code is already part of that broader discussion.