Brief IA

Anthropic Revolutionizes LLM-Anthropic 0.26 with Claude 5

💼 Business & Startups·Tom Levy·

Anthropic Revolutionizes LLM-Anthropic 0.26 with Claude 5

Anthropic Revolutionizes LLM-Anthropic 0.26 with Claude 5
Key Takeaways
1Version 0.26 of llm-anthropic introduces the models Claude-fable-5, Claude-sonnet-5, and Claude-opus-5.
2New server-side tools, such as WebSearch and CodeExecution, are now available through the LLM -T interface.
3The update requires llm>=0.32, enabling the streaming of tool results as typed events.
💡Why it mattersThis update significantly enhances the reasoning and interaction capabilities of the Claude models, increasing their utility for developers.
Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

Anthropic Revolutionizes llm-anthropic 0.26 with Claude 5

Access to Claude Models

The new version 0.26 of llm-anthropic marks a significant advancement with the integration of models from the Claude series. This update, made possible by LLM 0.32, offers unprecedented features.

The new models, namely claude-fable-5, claude-sonnet-5, and claude-opus-5, enhance Anthropic's offerings. Additionally, server-side tools such as WebSearch, WebFetch, CodeExecution, and AnthropicMCP are now accessible via the -T interface of LLM or through Python tools. The previous -o web_search* options have been replaced by -T WebSearch.

Technical Improvements

The update requires a version of llm>=0.32, allowing for reasoning, tool calls, and tool results to be broadcast as typed events. Reasoning for LLM CLI prompts is now visible in standard error, unless the --hide-reasoning/-R option is used.

The simplification of thinking has been introduced with levels of thinking_effort (low, medium, high, very high, or max). The Claude 5 models are configured to think by default; however, the -o thinking 0 option can disable this feature for Sonnet 5 and Opus 5, while Fable 5 retains this capability. The -R/--hide-reasoning option now allows for omitting reasoning from responses and logs. The options thinking_budget, thinking_display, and thinking_adaptive have been removed.

Release Date

This version was released on August 4, 2026 at 10 PM.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.