LangChain shipped langchain-anthropic==1.7.5 and langchain-core==1.6.6 on 2026-09-29, landing #40882: Claude Sonnet 5.5 model profile and capability discovery; reject unsupported forced tool_choice/thinking; allow mid-conversation system/tool changes; preserve computer-toolset namespaces, encrypted advisor blocks, and refusal details through streaming and replay. Invalid tool calls can fall back to unforced structured output or explicit method="json_schema".

Key Takeaways

  • ✓Releases langchain-anthropic==1.7.5 + langchain-core==1.6.6
  • ✓Core PR #40882 merged 2026-09-29T13:05Z
  • ✓Validates thinking/tool_choice; mid-conversation system/tool changes; preserves streamed provider blocks
  • ✓Forced tool-choice fallback or explicit method="json_schema"
  • ✓Install: pip install -U "langchain-anthropic==1.7.5" "langchain-core==1.6.6"; docs Anthropic provider
🧭

Finished reading? Explore benchmark rankings & pricing

Real-world SWE-bench scores & $20/mo vs API cost break-even calculator

🏆 Leaderboard→💰 Compare Plans↗
🔗

Project Links & Resources

Direct Access
Direct access to official project resources and documentation
🐙
langchain-anthropic==1.7.5 release
github.com/langchain-ai/langchain/releases/tag/langchain-anthropic==1.7.5
↗
🐙
langchain-core==1.6.6 release
github.com/langchain-ai/langchain/releases/tag/langchain-core==1.6.6
↗
🐙
Sonnet 5.5 compatibility PR #40882
github.com/langchain-ai/langchain/pull/40882
↗
📖
LangChain Anthropic provider docs
docs.langchain.com/oss/python/integrations/providers/anthropic
↗
📖
PyPI langchain-anthropic 1.7.5
pypi.org/project/langchain-anthropic/1.7.5/
↗
📖
PyPI langchain-core 1.6.6
pypi.org/project/langchain-core/1.6.6/
↗
🌐
ChatAnthropic reference
reference.langchain.com/python/integrations/langchain_anthropic/
↗
🔬

In-Depth Technical Analysis

Background

Clients already default to Claude Sonnet 5.5, but older langchain-anthropic builds reject unsupported thinking/tool_choice combos and drop streamed toolset/refusal metadata. Teams need library-level compatibility, not app-level hacks.

What shipped

#40882 lands in langchain-anthropic==1.7.5 with langchain-core==1.6.6: Sonnet 5.5 profile + capability discovery, validation for thinking/tool_choice, mid-conversation system/tool changes, and preservation of computer-toolset namespaces, encrypted advisor blocks, and refusals through stream/replay.

Benchmarks

No model leaderboard in the notes—this is an integration release. Compare tool-loop failure and thinking rejection rates on your eval set for 1.7.4→1.7.5.

Get started

pip install -U "langchain-anthropic==1.7.5" "langchain-core==1.6.6"; docs at the Anthropic provider page.