Claude Opus 5 Launches: Anthropic Brings Frontier Coding and Agentic AI at Unchanged Pricing

Anthropic released Claude Opus 5 on 24 July 2026, replacing Claude Opus 4.8 as the company’s Opus-tier flagship. The new model retains the same pricing as its predecessor—$5 per million input tokens and $25 per million output tokens—while delivering performance that Anthropic says approaches Claude Fable 5 at roughly half the cost. Opus 5 is now the default model on Claude Max and the strongest model available on Claude Pro.

The launch comes as Indian AI developers, startups, and enterprise teams increasingly evaluate frontier models for code generation, automation, and agentic workflows. Opus 5’s combination of improved coding benchmarks, safety guardrails, and unchanged pricing could make it a strong candidate for teams that need high-capability models without a cost jump.

Model specifications and context window

The model ID is claude-opus-5. It supports a 1 million token context window as both default and maximum, with no smaller variant. Maximum output is 128,000 tokens on the synchronous Messages API, and up to 300,000 tokens on the Message Batches API when using the output-300k-2026-03-24 beta header. The minimum cacheable prompt has been reduced to 512 tokens, down from 1,024, which could lower latency and cost for repeated prompt segments.

Benchmark performance across coding, agentic tasks, and reasoning

On FrontierBench v0.1, a 74-task successor to Terminal-Bench 2.1, Opus 5 scored 43.3% at max effort, compared to 18.7% for Opus 4.8, 33.7% for Fable 5, and 37.5% for GPT-5.6 Sol. At xhigh effort, Opus 5 reached 44.4% mean reward, its best result.

Coding benchmarks show significant gains. Opus 5 scored 96.0% on SWE-bench Verified and 79.2% on SWE-bench Pro, narrowly behind Fable 5’s 80.0% on Pro. On SWE-bench Multimodal, the jump was larger: from 38.4% (Opus 4.8) to 59.4%.

Agentic capabilities also improved. On OSWorld 2.0, Opus 5 reached 70.57% against 55.7% for Opus 4.8. On Zapier AutomationBench, it scored 26.0%, compared to 17.0% for Opus 4.8 and 17.4% for Fable 5. At medium effort, it still achieved 24% at $0.89 per task.

In reasoning, Anthropic prompted Opus 5 on all six IMO 2026 problems without tools or an agent harness. A three-model judge panel scored all 24 generated solutions as correct. Human experts independently graded one pre-specified solution per problem at 7/7, giving a final 42/42—gold-medal level, above the 29/42 cutoff.

The ARC Prize Foundation reported a verified 30.16% on ARC-AGI-3 at high effort, roughly four times the best previously reported leaderboard score. GPT-5.6 Sol reached 7.78% and Opus 4.8 reached 1.52%.

Key facts: Claude Opus 5 at a glance

Metric	Value
Pricing (input / output)	$5 / $25 per million tokens
Context window	1 million tokens (default and max)
FrontierBench v0.1 (max effort)	3%
SWE-bench Verified	0%
SWE-bench Pro	2%
OSWorld 2.0	57%
Zapier AutomationBench	0%

Safety improvements and reduced refusals

Anthropic’s system card and safety evaluations show that Opus 5’s classifiers flagged and refused 5% of API calls across 4% of trials, compared to Fable 5’s 42% of calls across 26% of trials. The company estimates that Opus 5’s safety classifiers will intervene around 85% less often than on Fable 5.

On the Gray Swan indirect prompt injection benchmark, attacker success within 15 attempts fell from 5.5% on Opus 4.8 to 2.0% on Opus 5. In browser environments run through Claude Cowork, attack success dropped from 31.5% on Opus 4.8 to 3.70% with no safeguards applied. With auto mode enabled, it reached 0% across all 129 environments.

Cybersecurity evaluations show that Opus 5’s capability to find vulnerabilities is nearly on par with Claude Mythos 5, but its ability to exploit them is substantially lower. Under the Responsible Scaling Policy, Anthropic treats Opus 5 as having CB-1 capabilities but not CB-2, and applies the same ASL-3 protections used for Opus 4.8. The AI R&D threshold is not crossed.

What this means for Indian AI developers and enterprises

For Indian teams building AI-powered coding assistants, automation tools, or agentic workflows, Opus 5 offers a meaningful upgrade without increasing costs. The improvements in coding benchmarks, tool use, and safety could reduce the need for multiple model tiers or extensive prompt engineering. The reduced context cache threshold (512 tokens) may also lower operational costs for applications that repeat prompts.

However, Indian developers should note that Opus 5’s safety classifiers are tuned to block exploit generation and certain cybersecurity tasks. Teams working on penetration testing or binary-level security analysis will need to apply for Anthropic’s Cyber Verification Program. The model’s strong performance on agentic benchmarks like OSWorld and Zapier AutomationBench suggests it is well-suited for real-world automation tasks common in Indian SaaS and B2B startups.

Source: MarkTechPost – “Meet the New Claude Opus 5: Frontier-Class Agentic Coding and Computer Use at Unchanged Opus Pricing” – https://www.marktechpost.com/2026/07/24/meet-the-new-claude-opus-5-frontier-class-agentic-coding-and-computer-use-at-unchanged-opus-pricing/