cd /news/developer-tools/llm-anthropic-0-26-adds-claude-5-and… · home topics developer-tools article
[ARTICLE · art-87157] src=letsdatascience.com ↗ pub= topic=developer-tools verified=true sentiment=· neutral

llm-anthropic 0.26 Adds Claude 5 and Server-Side Tools

Simon Willison released llm-anthropic 0.26 on August 4, 2026, adding support for Claude Fable 5, Sonnet 5, and Opus 5 model IDs to the open-source plugin that connects the LLM command-line tool and Python library to Anthropic's API. The release also exposes Anthropic's server-side WebSearch, WebFetch, CodeExecution, and AnthropicMCP tools through LLM's common tool interface, requires LLM 0.32 or newer, and replaces older extended-thinking options with new thinking and thinking_effort controls.

read3 min views1 publishedAug 5, 2026
llm-anthropic 0.26 Adds Claude 5 and Server-Side Tools
Image: Letsdatascience (auto-discovered)

Simon Willison released llm-anthropic 0.26 on August 4, adding Claude Fable 5, Sonnet 5 and Opus 5 model IDs to the LLM plugin. The release also exposes Anthropic server-side WebSearch, WebFetch, CodeExecution and MCP tools through LLM's tool interface, upgrades the plugin to LLM 0.32, and replaces older thinking options with the new thinking and thinking_effort controls.

Simon Willison released llm-anthropic 0.26 on August 4, 2026. The open-source plugin connects the LLM command-line tool and Python library to Anthropic's API, and the new version updates both the model catalog and the way developers invoke Anthropic-specific capabilities.

Claude 5 model IDs arrive in the plugin

Version 0.26 registers claude-fable-5, claude-sonnet-5 and claude-opus-5. That lets LLM users select those API model IDs through the same command-line and Python interfaces used for other supported models. The release does not introduce the models themselves; it adds support for them to this particular plugin.

The package now requires LLM 0.32 or newer. As a result, reasoning, tool calls, tool results and results from server-side tools can stream as typed events. Reasoning text is shown on standard error by default in the command-line interface, while --hide-reasoning or -R suppresses it.

Server-side tools use the common tool interface

The plugin exposes Anthropic's WebSearch, WebFetch, CodeExecution and AnthropicMCP server-side tools through LLM's -T option or the Python tools= argument. The former web_search options were removed in favor of the tool interface, so scripts that use those older switches need to migrate.

Extended-thinking controls are also simpler. The supported options are now thinking and thinking_effort, with effort levels from low through max. Older thinking_budget, thinking_display and thinking_adaptive options have been removed. Claude 5 models think by default under the plugin's new behavior; thinking can be disabled for Sonnet 5 and Opus 5, while Fable 5 always uses it.

For practitioners, the release is chiefly a compatibility and migration update. Upgrade LLM and the plugin together, replace removed web-search and thinking options, and test any code that consumes streamed responses because the 0.32 stack now emits structured event types rather than treating every chunk as plain text.

Key Points #

  • 1llm-anthropic 0.26 registers Claude Fable 5, Sonnet 5 and Opus 5 model IDs without presenting the plugin release as the launch of those models.
  • 2Anthropic server-side web search, web fetch, code execution and MCP capabilities now use LLM's common tool interface instead of the removed web_search options.
  • 3The plugin requires LLM 0.32 and replaces older extended-thinking settings with thinking and thinking_effort controls while streaming structured reasoning and tool events.

Scoring Rationale #

The plugin release gives LLM users immediate access to current Claude model IDs and a unified path for Anthropic server-side tools, with concrete migration work for removed options. Its impact is meaningful for developers using this integration but narrower than the underlying model launches.

Sources #

Primary source and supporting public references used for this report.

Practice interview problems based on real data

1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.

Try 250 free problems

── more in #developer-tools 4 stories · sorted by recency
gist.github.com · · #developer-tools
ECC.md
── more on @simon willison 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/llm-anthropic-0-26-a…] indexed:0 read:3min 2026-08-05 ·