Anthropic’s Claude Sonnet 5 Bridges Performance Gap with Opus Models

Anthropic has launched Claude Sonnet 5, its latest AI model, which demonstrates significant advancements in performance, narrowing the gap with its higher-priced Opus series. The new model, available at an introductory price, is positioned as Anthropic’s most “agentic” Sonnet yet, capable of building plans, utilizing tools like browsers and terminals, and executing tasks autonomously.

This release is particularly relevant for Indian businesses and developers looking for powerful yet cost-effective AI solutions. Sonnet 5’s improved capabilities could enable more sophisticated automation and intelligent applications across various sectors in India, from customer service to complex data analysis.

Key Performance Improvements

Anthropic’s benchmarks highlight Sonnet 5’s substantial upgrades over its predecessor, Sonnet 4.6. The model shows improvements across all tested categories, notably in agentic coding and multidisciplinary reasoning. For instance, on SWE-bench Pro for agentic coding, Sonnet 5 achieved 63.2%, an increase from Sonnet 4.6’s 58.1%. While Opus 4.8 still leads at 69.2%, the gap has noticeably shrunk.

In multidisciplinary reasoning (Humanity’s Last Exam), Sonnet 5, with tools, reached 57.4%, closely trailing Opus 4.8’s 57.9%. More remarkably, on the GDPval-AA v2 knowledge work benchmark, which assesses real-world knowledge tasks, Sonnet 5 surpassed Opus 4.8 with a score of 1,618 against Opus’s 1,615. This indicates Sonnet 5’s strong performance in practical knowledge-intensive applications.

Enhanced Agentic Capabilities

Anthropic emphasizes Sonnet 5’s enhanced agentic capabilities, meaning the model can undertake more complex, multi-step tasks independently. Early-access partners have reported that Sonnet 5 handles search tasks with greater efficacy and autonomy than previous versions. This ‘agentic’ quality is crucial for developing AI systems that can operate with minimal human intervention, a goal for many enterprises in India seeking to optimize operations.

The model’s ability to “build plans” and “grab tools” signifies a step towards more autonomous AI agents, which could be transformative for workflow automation, research, and development in various industries.

Cybersecurity and Safety Considerations

The launch of Sonnet 5 occurs amidst ongoing discussions about AI safety and cybersecurity. Anthropic has addressed these concerns by stating that Sonnet 5 was not trained on cybersecurity tasks and scores significantly lower than models like Mythos 5 and Opus 4.8 in tests for risky capabilities, such as generating software exploits.

Despite a slight increase in risky task scores compared to its direct predecessor, Anthropic has implemented default cyber safeguards in Sonnet 5. These protections are designed to flag and block risky cyber usage in real-time, aligning with the secure environment established for Claude Opus 4.7 and 4.8. The company has also improved the model’s ability to resist malicious requests and prompt injection attacks, reducing hallucinations and sycophantic behavior.

Availability and Pricing

Claude Sonnet 5 is now live across all Anthropic plans, serving as the new default for Free and Pro users. Max, Team, and Enterprise subscribers also have access. Developers can integrate it via Claude Code and the Claude Platform, with the API name “claude-sonnet-5”.

The model has a training cutoff of January 2026 and features a one-million-token context window. Until August 31, 2026, Anthropic is offering an introductory price of $2 per million input tokens and $10 per million output tokens. Following this period, prices will adjust to $3 and $15, matching the previous Sonnet models. However, Anthropic notes that due to its increased agentic nature, Sonnet 5 might consume more tokens per task, potentially leading to higher overall costs despite competitive per-token rates.

Key facts

Feature	Detail
Model Name	Claude Sonnet 5
Key Improvement	Enhanced agentic capabilities, closing gap with Opus series
Benchmark Highlight	Outperformed Opus 4.8 in GDPval-AA v2 knowledge work test
Introductory Price	$2/million input tokens, $10/million output tokens (until Aug 31, 2026)

Implications for India

For the Indian market, Claude Sonnet 5 represents a significant opportunity. Its balance of advanced performance and competitive pricing makes it an attractive option for startups, SMEs, and larger enterprises looking to integrate sophisticated AI into their operations without the higher costs associated with top-tier models like Opus. The enhanced agentic capabilities could drive innovation in areas such as intelligent automation for customer support, content generation, data analysis, and software development, fostering digital transformation across various sectors in India.

Source: The Decoder – https://the-decoder.com/anthropics-new-claude-sonnet-5-closes-the-gap-to-the-pricier-opus-model-series/