Anthropic released Claude Opus 5 on July 24, 2026, and the model is available the same day across all of Anthropic’s platforms. The company positions Opus 5 as a more thoughtful and proactive model that approaches the frontier intelligence of Claude Fable 5 at roughly half the cost per task. Opus 5 is now the default model on Claude Max and the strongest model available on Claude Pro.
What does Claude Opus 5 do better than Opus 4.8?
Anthropic says Opus 5 delivers improved performance over Opus 4.8 at the same price, with three areas called out as the biggest gains:
- Stronger verification and iteration on long-running tasks.
- Better visual output generation, including interactive illustrations and data visualizations.
- Improved scientific research capability, particularly in life sciences.
How does Opus 5 score on coding and agent evaluations?
On coding and knowledge-work benchmarks, Anthropic reports Opus 5 leads, though it still trails Mythos 5 on cybersecurity tasks.
- Frontier-Bench v0.1: Opus 5 surpasses every other model and more than doubles Opus 4.8’s performance at a lower cost per task.
- CursorBench 3.2: At max effort, Opus 5 lands within 0.5% of Fable 5’s peak score at half the cost, and beats every other model on high, xhigh, and max effort settings.
- ARC-AGI 3: Opus 5 scores roughly three times higher than the next-best model on this evaluation of novel problem solving.
- Zapier AutomationBench: Pass rate is about 1.5 times the next-best model at the same cost, and even Opus 5’s lowest effort setting passes more tasks than any other model.
- OSWorld 2.0: A computer-use benchmark where Opus 5 outperforms every other model at any given cost, surpassing Fable 5’s best result at about one-third the cost.
- Coding Agent Index: Listed by Anthropic as a top result for Opus 5.
According to the announcement, Opus 5 is also Anthropic’s best and most cost-efficient model on related evaluations including GDPval-AA v2, HLE AutomationBench, and DeepSearchQA.
How much better is Opus 5 at scientific research?
Anthropic says Opus 5 outperforms Opus 4.8 on every life sciences evaluation it ran, covering structural biology, organic chemistry, and bioinformatics. The largest reported gains:
- Organic chemistry: 10.2 percentage points higher than Opus 4.8 on an internal benchmark involving inferring molecular structures from spectroscopy data.
- Protein-related tasks: 7.7 percentage points higher than Opus 4.8 on work that predicts how sequence variations affect protein function.
How is Opus 5 priced?
Pricing matches Opus 4.8 at 5 dollars per million input tokens and 25 dollars per million output tokens. A faster mode runs about 2.5 times faster at double the base price. Customers can adjust an effort setting to trade off intelligence against token spend.
What early-access customers are saying
Anthropic included brief testimonials from companies testing Opus 5 before launch:
- Devin (Scott Wu, CEO): On FrontierCode 1.1, Opus 5 approaches Fable-level performance at half the cost, with particular strength on difficult debugging and root-cause analysis.
- Cursor (Sualeh Asif, Co-Founder): Near Fable 5 intelligence at Opus speed and cost, within 0.5% on CursorBench with similar behaviors.
- Zapier (Wade Foster, CEO): Opus 5 topped Zapier’s AutomationBench leaderboard without spending more tokens than prior Claude models, hitting 100% on a churn-prevention workflow where previous models did not pass.
- Genomics analysis firm (Alfredo Andere, CEO): Opus 5 behaves more like a careful scientist, using appropriate statistical tests and cross-checking its own results.
- Lovable (Fabian Hedin, Co-Founder): Opus 5 came out ahead of every model in its family on internal evaluations, up 22% over Opus 4.7 on the hardest agentic coding tasks, with steadier run-to-run results.
- Box (Ben Kus, CTO): Opus 5 outperforms Opus 4.8 by 8%, with an 11% gain in data analysis workflows and a 17% gain in due diligence workflows.
- Trading firm (Richard Pham, Evals and Product Lead): On financial-modeling tasks, Opus 5 averaged 9 percentage points higher accuracy with a third fewer turns and tool calls, and 60% less time.
- JetBrains (Denis Shiryaev, Head of AI in IDE): Opus 5 plans more carefully, catches its own logical faults during planning, and reasons about why an answer is right rather than just whether it works.
- AWS Kiro (Deepak Singh, VP of Agentic AI): A strong agentic coding model for long-running, multi-step work that holds context and pins down requirements better than Opus 4.8.
Several other teams from companies including LovedByHumans, Box, JetBrains, and unnamed trading and engineering firms described gains in consistency, numerical reasoning, and code-review judgment.
How does Opus 5 handle verification and agentic work?
Anthropic shared three specific examples of Opus 5 verifying and iterating on its own work during evaluations and early-access testing:
- On one Frontier-Bench task, Opus 5 was given a drawing of a machine part and asked to rebuild it as a 3D FreeCAD model, with no way to directly view the drawing. The model wrote its own computer vision pipeline to extract geometry from raw pixels and reconstructed the part repeatedly. No competing model solved 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 had 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 could not complete the task at all, even with extensive plans from the engineer. Opus 5 also built its own test because no live feed was available to validate against.
What did Anthropic find in alignment and safety testing?
Anthropic’s automated behavioral audit ranked Opus 5 as the company’s most aligned model to date, scoring 2.3 on overall misaligned behavior, the lowest of recent Claude models. Opus 5 adheres to Claude’s Constitution better than Opus 4.8, Sonnet 5, or Fable 5, shows the lowest rates of deceptive behavior, and is the least susceptible to being tricked into misuse.
On dual-use capabilities, Anthropic reports that Opus 5 does not advance the frontier in risky domains. Evaluations run with private-sector and government partners found the model still trails Mythos 5 in biology research and offensive cybersecurity. As with Opus 4.8, Anthropic avoided training Opus 5 on cyber tasks, though general capability gains brought it close to Mythos 5 at finding vulnerabilities. On exploitation of those vulnerabilities, Opus 5 remains substantially behind Mythos 5, as shown on the OSS-Fuzz evaluation.
FAQ
When was Claude Opus 5 released?
Anthropic announced Claude Opus 5 on July 24, 2026, and the model became available the same day across all of Anthropic’s platforms.
How much does Claude Opus 5 cost?
Claude Opus 5 is priced at 5 dollars per million input tokens and 25 dollars per million output tokens, the same pricing as Opus 4.8. A fast mode runs about 2.5 times faster at double the base price.
How does Opus 5 compare to Claude Fable 5?
Anthropic positions Opus 5 as approaching the frontier intelligence of Fable 5 at roughly half the cost per task. On specific benchmarks, Opus 5 lands within 0.5% of Fable 5’s peak score on CursorBench 3.2 at max effort and surpasses Fable 5’s best result on OSWorld 2.0 at about one-third the cost.
Related coverage
- Anthropic keeps Claude Fable 5 inside Max and Team Premium with permanent usage caps
- White House Accuses Moonshot AI of Distilling Anthropic’s Claude for Kimi K3
Want help with this in Myrtle Beach or Long Island? Talk to us.
NewSunSEO
Grow in Google and AI Search
AI-SEO, local search, paid ads and websites from one team, with offices on Long Island and in Myrtle Beach.
- AI-SEO
- Local AI-SEO
- AI Consulting
- PPC
- Web Design
- Web Development
- Long Island Office(631) 320-1700
- Myrtle Beach Office(843) 492-7131
Request a Free Proposal
Tell us about your business and where you want to grow.
We will reply by email to talk through your goals and what a proposal would include.
We use your details only to reply to you and send what you asked for. See our Privacy Policy.
