What Changed: A Daily-Driver Model That Approaches Frontier Intelligence
On July 24, 2026, Anthropic released Claude Opus 5, positioning it as a “thoughtful and proactive model that comes close to the frontier intelligence of Claude Fable 5 at half the price.” For product builders, this isn’t just another spec bump—it’s a practical option to get near-top-tier performance without the top-tier cost.
According to Anthropic’s official announcement, Opus 5 is the new state-of-the-art on coding and knowledge work evaluations like Frontier-Bench and GDPval-AA, though it remains behind Mythos 5 on cybersecurity tasks. It’s designed for everyday use, works more efficiently than other models, and is the new default on Claude Max and the strongest model on Claude Pro.
Performance and Cost-Effectiveness: More Than Just Cheaper
Anthropic’s data shows Opus 5 delivers greatly improved performance for the same cost as its predecessor, Opus 4.8. The model’s effort setting lets customers optimize for intelligence or conserve tokens for faster, cheaper results.
On Frontier-Bench v0.1, Opus 5 surpasses all other models and more than doubles Opus 4.8’s performance at a lower cost per task. On CursorBench 3.2, at max effort, it performs within 0.5% of Fable 5’s peak score but at half the cost per task. It also achieves greater performance at a given cost than all other models on high, xhigh, and max effort.
Knowledge work and problem-solving show similar gains:
- On ARC-AGI 3, which tests novel problem-solving, Opus 5’s score is three times as high as the next-best model.
- On Zapier AutomationBench, which measures end-to-end business task completion, Opus 5’s pass rate is around 1.5× the next-best model for the same cost per task. Even at its lowest effort setting, it passes more tasks than any other model.
- On OSWorld 2.0, a computer use benchmark, Opus 5 outperforms every other model at any given cost, surpassing Fable 5’s best result at just over a third of the cost.
For scientific research, Opus 5 is a meaningful improvement over Opus 4.8 across all life sciences evaluations, including structural biology, organic chemistry, and bioinformatics. It scores 10.2 percentage points higher on organic chemistry tasks (like inferring molecular structures from spectroscopy data) and 7.7 percentage points higher on protein-related tasks (like predicting how sequence variations affect function).
A legal tech company reported that Opus 5 maintained similar quality while generating 26% fewer tokens on average compared to Opus 4.8 at max reasoning. This token efficiency is a key part of the cost story.
The Real Highlight: Judgment and Self-Verification
Multiple early-access customers used the word “judgment” to describe Opus 5. One engineer described a rearchitecting session where Opus 5 pushed back on a proposed design but didn’t fold when the engineer insisted. Instead, it explained what was valuable in the idea, narrowed its objection to a single design question, and proposed a compromise that kept the good part while fixing the flaw. That kind of behavior lets developers trust it with less oversight.
Opus 5 is also much stronger at verifying its work and iterating carefully. Anthropic and users found several examples:
- On a Frontier-Bench task, Opus 5 was given a drawing of a machine part and asked to write code to rebuild it as a 3D FreeCAD model, but with no way to directly view the drawing. Opus 5 wrote its own computer vision pipeline to pull geometry from raw pixels and reconstructed the part successfully, repeatedly. No competing model could solve it after five attempts.
- Given a real bug in a popular open-source package manager, Opus 5 found the root cause and fixed an edge case the community’s patch missed. A competing model fixed only the surface symptom and reported the bug resolved.
- An engineer at a trading firm used Opus 5 to build a market data feed for a new exchange in a single session. Previous models couldn’t complete the task at all. With no live feed to validate against, Opus 5 built its own test harness to check that its code parsed the exchange’s data correctly.
Customers echoed this thoroughness. A frontend benchmark showed Opus 5 opening pages in a browser at desktop and phone widths, catching a product hidden below the mobile fold and an off-screen checkout button, and fixing both before handing the work back. A monitoring agent built with Opus 5 flagged a potential anomaly, re-checked its assumption against production, found the signal benign, wrote the correction into its memory, and retired its monitoring queries on its own.
Practical Use Cases: From Code to Long-Horizon Work
For product builders, the value isn’t just “stronger” but “more reliable.” Lovable reported Opus 5 is up 22% over Opus 4.7 on their hardest agentic coding tasks, with far less variance run to run. That consistency is critical for products that depend on AI generating consistent results.
Opus 5 also excels at long-horizon tasks. One customer gave it a chief-of-staff role over dev environments: it built its own monitor, drove each box, and pulled the human in only for judgment calls. Another used it for financial research, averaging 9 percentage points higher accuracy with a third fewer turns and tool calls and 60% less time.
In agentic workflows, Opus 5 handles larger changes. A customer noted it made large-scale changes across a codebase, adapting to feedback throughout an agentic workflow and explaining its reasoning more clearly than any model they’d used. It handled work they’d normally break into much smaller pieces.
For code review, Opus 5 writes clean, tight diffs with no dead code and is a stronger hazard spotter on subtle, codebase-specific issues. It verifies branches, checks templates, and thinks through test implications before handing off a PR, avoiding the rushed mistakes of older models.
Limitations and Trade-Offs
Opus 5 is not a universal upgrade. Anthropic acknowledges it remains behind Mythos 5 on cybersecurity tasks, particularly in exploiting vulnerabilities. On OSS-Fuzz, Opus 5 is close to Mythos 5 at identifying vulnerabilities but considerably less successful at developing exploits. It also doesn’t advance the frontier in risky dual-use capabilities like biology research.
Official benchmarks come from specific evaluations; real-world results vary by task type. Product builders should test in their own workflows, paying attention to the effort setting. Anthropic offers different effort levels to balance intelligence and token cost.
Safeguards also differ. Opus 5’s cyber classifiers are less restrictive than Fable 5’s, allowing vulnerability finding but blocking binary-based scanning, penetration testing, and exploit generation. Flagged requests fall back to Opus 4.8 by default, and automatic fallbacks are available on the API.
Concrete Takeaway: Test on Hard, Vague Tasks
Claude Opus 5 is available today on all platforms at $5 per million input tokens and $25 per million output tokens (same as Opus 4.8). Developers can start with claude-opus-5 on the Claude API. It’s also offered in Fast mode at about 2.5× default speed, at twice the base price.
For developers, the most direct step is to try Opus 5 in integrated environments like Cursor, Devin, and Kiro, and compare it with Opus 4.8 on difficult, ambiguous tasks—multiple customers noted this is where Opus 5 improves most.
If you value cost efficiency, Opus 5’s low effort setting may suffice for daily tasks. For complex, multi-step projects, increase the effort level. Ultimately, whether Opus 5’s judgment translates into a competitive advantage for your product depends on how you design workflows to leverage its self-verification and long-horizon planning strengths.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.