TLDRocket
Sign in

Claude Sonnet 5

Model ● Covered in 11 stories Compare ⇄ + Follow

Claude Sonnet 5 is Anthropic's latest model released as the default offering for Free and Pro users, designed to handle autonomous tasks like planning and tool use with performance approaching the more expensive Opus 4.8 model. The model is priced at $2 per million input tokens and $10 per million output tokens through August 2026, though recent analysis indicates its tokenizer produces approximately 30% more tokens than previous Claude versions and 73% more than GPT's tokenizer, making actual per-task costs higher than list prices suggest. Despite lower per-token pricing, Sonnet 5 has shown mixed results in practice, consuming 10-12x more tokens on common agent tasks and sometimes producing lower quality output than Opus 4.8, while also exhibiting regressions in tool-calling reliability in some agentic contexts.

Updated 3 August 2026

Specifications

No specifications recorded yet.

Latest developments

Are AI labs pelicanmaxxing? Simon Willison · 1 week ago · 44 Last Week in AI #251 - Mythos Back, Sonnet 5, Etched, LongCat Last Week in AI · 1 week ago · 12 [AINews] Kimi K3 2.8T-A50B: the largest open model ever released; Opus 4.8-class at Sonnet 5 pricing Latent Space · 2 weeks ago · 19 Introducing Claude Sonnet 5 Anthropic News · 43 The Same TypeScript Costs 73% More Tokens on Claude Than GPT TLDR Dev · 2 weeks ago · 4 Not all model upgrades are upgrades TLDR Dev · 3 weeks ago · 43
Price per 1M tokens is meaningless TLDR · 3 weeks ago · 48 Better Models: Worse Tools TLDR · 4 weeks ago · 35

Timeline

Month Quarter Year

Q3 2026

Study finds no evidence of AI labs optimizing models for pelican-on-bicycle image generation Research publication

Multiple AI model releases and regulatory developments across OpenAI, Anthropic, and competitors Other

Moonshot AI releases Kimi K3, a 2.8 trillion parameter open-weight model achieving frontier-class performance Open source release

Claude Sonnet 5 Release and Tokenizer Efficiency Comparison with Competitors Model release

Analysis reveals Claude Sonnet 5 has higher effective costs and lower quality than predecessor despite lower per-token pricing Feature update

Anthropic redeployed Fable 5 AI model with enhanced safety restrictions following government export controls Feature update

Relationships

Products & technology

  • Anthropic develops this model · 10 sources

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads 60+ sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.