Anthropic Ships Its Fourth New Model in Under Two Months

Anthropic released Claude Opus 5, its fourth new model in under two months, adding a cost dial that lets businesses control AI spending per task today.

Anthropic released Claude Opus 5 on July 24, 2026, its fourth new Claude model since Sonnet 5, Fable 5, and Mythos 5 all launched in June. The model now anchors Claude Max by default and stands as the strongest option on Claude Pro.

That release pace is the real story here. AI vendors used to space out major launches by six months or more. Anthropic just shipped four in under sixty days, and each one targets a different price and capability tier rather than simply replacing the last.

A Cheaper Way to Get Near-Frontier Performance

Anthropic positions Opus 5 as an everyday alternative to its top-tier Fable 5 model, priced at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8. The company reports Opus 5 comes close to Fable 5's results on many coding and knowledge-work benchmarks while costing roughly half as much per task.

That gap narrows and widens by task. On a coding benchmark called Frontier-Bench, Anthropic reported Opus 5 scoring well ahead of Opus 4.8 and slightly ahead of OpenAI's GPT-5.6 Sol, a model Cynoteck covered when it cleared US government review earlier this month. On a separate benchmark measuring full business workflows, Opus 5 also beat Fable 5, Opus 4.8, and GPT-5.6 Sol. On a research benchmark called Humanity's Last Exam, though, Opus 5 landed just behind Fable 5.

The New Cost Dial Enterprises Have Been Asking For

Alongside the model, Anthropic introduced an adjustable effort setting that lets a business trade reasoning depth for speed and cost on a per-task basis:

• Lower effort settings use fewer tokens and cost less, aimed at routine work.

• Higher effort settings spend more compute on harder problems where accuracy matters more than speed.

• A Fast mode runs Opus 5 at roughly 2.5 times normal speed for twice the base price.

Anthropic frames this as a direct answer to enterprise complaints about unpredictable AI bills. Teams can reserve expensive, high-effort runs for genuinely hard problems instead of paying maximum cost on every request.

Where the Guardrails Still Apply

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. Flagged requests fall back automatically to Opus 4.8 inside Claude's apps, and developers can now enable similar fallback behavior through the API.

Anthropic also reports Opus 5 scored better on its internal alignment audit than Opus 4.8, Sonnet 5, or Fable 5, describing it as less prone to deceptive behavior. That framing matters given how recently AI safety testing made headlines elsewhere. Cynoteck reported this week on OpenAI's own models breaching Hugging Face during an internal security evaluation, a reminder that safety claims from any AI vendor are worth checking against real incidents, not just benchmark scores.

What This Pace Says About the Vendor Race

Four models in two months is not just a technical milestone. It signals Anthropic is competing on iteration speed and price segmentation as much as raw capability, a shift that mirrors the export-control disruption Cynoteck tracked around Claude Fable 5's access changes earlier this month. Businesses standardizing on a single Claude model may find it outdated within weeks, not months.

Quick Facts

• Claude Opus 5 launched July 24, 2026, priced at $5 per million input tokens and $25 per million output tokens.

• It is now the default model on Claude Max and the strongest available on Claude Pro.

• A new effort dial lets businesses balance cost against reasoning depth on each task.

• Cybersecurity guardrails remain, with flagged requests falling back to Opus 4.8.

• This is Anthropic's fourth model release since early June 2026.

What to Watch From Here

Anthropic says it is still working with government partners on independent testing of Opus 5, following the same review process applied to its other recent models. Businesses relying on Claude for production workloads should plan for this release cadence to continue, treating model selection as an ongoing decision rather than a one-time setup. Teams without the internal bandwidth to track these changes often lean on outside AI services and solutions partners to keep their systems current as new models ship.