- Anthropic has unveiled Claude Sonnet 5, its latest AI model focused on delivering stronger agentic AI capabilities at a lower cost.
- While Sonnet 5 is not as capable as Anthropic’s flagship Claude Fable 5 or Mythos 5 models, it brings several improvements over Sonnet 4.6 and narrows the performance gap with Opus 4.8.
- In benchmark testing, Sonnet 5 scored 63.2% on an agentic coding benchmark, improving on Sonnet 4.6’s 58.1% while trailing Opus 4.8’s 69.2%.
- However, on a separate knowledge work benchmark, Sonnet 5 outperformed Opus 4.8, highlighting stronger performance in reasoning, research and complex knowledge-based tasks.
- According to early testers, Sonnet 5 is also better at completing complex, multi-step tasks that previously required manual intervention.
- The model was found to verify its own outputs more consistently, enabling it to execute end-to-end workflows with greater reliability.
- Anthropic says Sonnet 5 also delivers improved safety compared with its predecessor.
- The model demonstrates lower rates of hallucinations, deception and other undesirable behaviours, while showing stronger resistance to malicious prompts and prompt injection attacks.
- At launch, Sonnet 5 is priced at $2 per million input tokens and $10 per million output tokens through August 31.
- Beginning September 1, the input token price will increase to $3 per million, while the output token price will remain unchanged at $10 per million.
- Anthropic says Sonnet 5 is priced below Opus 4.8, OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro, although it remains more expensive than Gemini 3.5 Flash.
Anthropic has launched Claude Sonnet 5, an AI model designed to enhance agentic capabilities while offering a more affordable pricing structure. At $2 per million input tokens and $10 per million output tokens until August 31, the model aims to attract a broader user base.910
While Sonnet 5 is not as advanced as Anthropic's flagship models, it shows significant improvements over its predecessor, Sonnet 4.6, which scored 58.1% on the same benchmark. In contrast, Sonnet 5 achieved 63.2%, although it still lags behind Opus 4.8, which scored 69.2% on the agentic coding benchmark.23
Interestingly, Sonnet 5 outperformed Opus 4.8 on a separate knowledge work benchmark, demonstrating its strength in reasoning, research, and complex knowledge-based tasks. Early testers have noted that the model is particularly adept at handling complex, multi-step tasks that previously required manual intervention, indicating a significant leap in its operational capabilities.4
Starting September 1, the input token price will rise to $3 per million, while the output token price will remain at $10 per million, reflecting the company's strategy to balance affordability with performance enhancements.
“Anthropic has unveiled Claude Sonnet 5, its latest AI model focused on delivering stronger agentic AI capabilities at a lower cost. The model is priced at $2 per million input tokens and $10 per million output tokens through August 31.”
Firstpost
