Anthropic's Claude Sonnet 5 Offers Higher Performance, Lower Costs in AI Race
Anthropic has launched Claude Sonnet 5, a mid-sized artificial intelligence model that delivers advanced autonomous capabilities at a significantly lower cost than its flagship systems. The release reflects a broader industry shift from chatbots to AI systems capable of independent planning, reasoning, and complex task completion.
According to the company, Claude Sonnet 5 can create multi-step plans, use tools such as web browsers and computer terminals, and complete tasks with minimal human oversight. Anthropic says the model now performs at a level previously achieved only by much larger and more expensive AI systems just months ago.
The announcement follows similar moves by competitors. OpenAI recently introduced GPT-5.6 Sol with enhanced multi-agent capabilities, while Google positioned its Gemini 3.5 Flash model as an AI assistant for real-world tasks. As a result, autonomous or 'agentic' AI is becoming the new industry standard.
Anthropic's unique selling point for Claude Sonnet 5 is its balance of performance and cost. The company says the model delivers results close to its premium Claude Opus 4.8 model while being considerably cheaper to operate.
Pricing starts at $2 (approximately Rs 190) per million input tokens and $10 (approximately Rs 948) per million output tokens until August 31. After that, the cost will increase to $3 (approximately Rs 285) per million input tokens and $15 (approximately Rs 1,423) per million output tokens.
Even after the price revision, Anthropic says Sonnet 5 remains more affordable than OpenAI's GPT-5.5 and Google's Gemini 3.1 Pro, though it is priced above Google's Gemini 3.5 Flash.
Claude Sonnet 5 replaces earlier versions as the default model for both free and paid Claude users, making its upgraded capabilities immediately available across Anthropic's platform.
Performance improvements extend beyond pricing. Anthropic says the model performs better than its predecessor, Claude Sonnet 4.6, in reasoning, software development, tool use, and knowledge-intensive tasks. On agentic coding benchmarks, Sonnet 5 scored 63.2 per cent, compared with 58.1 per cent for Sonnet 4.6, while approaching the 69.2 per cent achieved by the larger Opus 4.8.
The company also claims Sonnet 5 is better at completing long-running tasks that earlier models often abandoned midway. It can review its own work without explicit instructions, making it more reliable for business automation and software workflows.
Early testers have reported noticeable improvements. Automation platform Zapier said the model successfully completed multi-step business workflows, such as updating Salesforce account data and sending enterprise announcements—tasks that previously stalled before completion.
