LangChain releases

4 items in the render window (newest 100 + all corrections; not this source's all-time total) · status healthy · last successful fetch 2026-11-21 19:16 UTC
Feed for this sourceRSS
WORTH KNOWING API LangChain releases · 2026-11-20 07:18 UTC

langchain-anthropic 1.7.12 requires langgraph 1.2.16 or newer

Do:Check your langgraph version before upgrading langchain-anthropic.
WHAT CHANGEDlangchain-anthropic 1.7.12, released 19 November 2026, raises the minimum langgraph dependency to version 1.2.16. The release also fixes async HTTP client isolation between models and preserves caller information on tool_search_tool_result blocks.
WHY IT MATTERSExisting installations with langgraph <1.2.16 will fail dependency resolution on upgrade. The async HTTP client fix prevents state leakage between concurrent model instances; the tool_search_tool_result fix changes how caller context propagates through tool invocations.
WHAT TO DOCheck your langgraph version before upgrading langchain-anthropic. If pinned below 1.2.16, update both together. Test tool-calling workflows if you rely on caller preservation.
Model-reported confidence 85% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## langchain-anthropic==1.7.12
+https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.7.12
+2026-11-19T23:00:21Z
+
+Changes since langchain-anthropic==1.7.11
+release(anthropic): 1.7.12 (#41966)
+feat(core): standardize finish reason on message-finish events (#41926)
+chore(langchain): require langgraph 1.2.16 or newer (#41959)
+fix(anthropic): preserve caller on tool_search_tool_result blocks (#41958)
+fix(anthropic): don't share the async HTTP client between models (#41946)
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.72, needed 0.5
triage reasoningNew dependency constraint and async client isolation fix affect runtime behavior and deployment requirements.
signalsversion bump with release notes · dependency version floor raised: langgraph 1.2.16 or newer · async HTTP client isolation fix (security/stability) · tool_search_tool_result caller preservation (behavior change)
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)
WORTH KNOWING API LangChain releases · 2026-11-19 20:39 UTC

langchain-openrouter 0.3.0 drops Python 3.10 support

Do:Check your Python version with `python --version`.
WHAT CHANGEDlangchain-openrouter 0.3.0 requires Python 3.11 minimum, up from Python 3.10. The minimum langchain-core version is now 1.8.3. Python 3.10 users will continue to resolve langchain-openrouter 0.2.x.
WHY IT MATTERSPython 3.10 environments cannot upgrade to langchain-openrouter 0.3.x. Dependency resolution will pin them to 0.2.x, which will eventually stop receiving updates.
WHAT TO DOCheck your Python version with `python --version`. If running Python 3.10, either upgrade to Python 3.11+ before adopting langchain-openrouter 0.3.x, or pin to langchain-openrouter<0.3 in your requirements.
Model-reported confidence 95% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## langchain-openrouter==0.3.0
+https://github.com/langchain-ai/langchain/releases/tag/langchain-openrouter%3D%3D0.3.0
+2026-11-19T15:03:21Z
+
+langchain-openrouter 0.3.0 drops support for Python 3.10, which has reached end-of-life. The minimum supported version is now Python 3.11, and the minimum langchain-core version is now 1.8.3. Users on Python 3.10 will continue to resolve langchain-openrouter 0.2.x.
+Changes since langchain-openrouter==0.2.9
+release(openrouter): 0.3.0 (#41957)
+chore(model-profiles): refresh model profile data (#41934)
+chore(model-profiles): refresh model profile data (#41917)
+chore(model-profiles): refresh model profile data (#41893)
+chore(model-profiles): refresh model profile data (#41861)
+chore(model-profiles): refresh model profile data (#41828)
+feat(partners): Remove EOL'd Python 3.10 from partner packages (#41268)
+chore(model-profiles): refresh model profile data (#41791)
+chore(model-profiles): refresh model profile data (#41781)
+chore(model-profiles): refresh model profile data (#41762)
+chore(model-profiles): refresh model profile data (#41745)
+chore(model-profiles): refresh model profile data (#41725)
+chore(model-profiles): refresh model profile data (#41706)
+chore(model-profiles): refresh model profile data (#41669)
+chore(model-profiles): refresh model profile data (#41651)
+chore(model-profiles): refresh model profile data (#41627)
+chore(model-profiles): refresh model profile data (#41600)
+chore(model-profiles): refresh model profile data (#41573)
+chore(model-profiles): refresh model profile data (#41556)
+chore(model-profiles): refresh model profile data (#41543)
+chore(model-profiles): refresh model profile data (#41520)
+chore(model-profiles): refresh model profile data (#41494)
+chore(model-profiles): refresh model profile data (#41469)
+chore(model-profiles): refresh model profile data (#41451)
+chore(model-profiles): refresh model profile data (#41432)
+chore(infra): relax third-party dependency upper bounds (#41407)
+chore(model-profiles): refresh model profile data (#41415)
+chore(model-profiles): refresh model profile data (#41354)
+chore(deps): bump certifi from 2026.2.25 to 2026.10.19 in /libs/partners/openrouter (#41378)
+chore(model-profiles): refresh model profile data (#41325)
+chore(model-profiles): refresh model profile data (#41298)
+chore(model-profiles): refresh model profile data (#41283)
+chore(model-profiles): refresh model profile data (#41265)
+fix: fix dependencies in some partner packages (#41266)
+chore(model-profiles): refresh model profile data (#41253)
+chore(model-profiles): refresh model profile data (#41242)
+chore(model-profiles): refresh model profile data (#41229)
+chore(model-profiles): refresh model profile data (#41215)
+chore(model-profiles): refresh model profile data (#41145)
+chore(model-profiles): refresh model profile data (#41117)
+chore(model-profiles): refresh model profile data (#41084)
+chore(model-profiles): refresh model profile data (#41045)
+chore(model-profiles): refresh model profile data (#41017)
+chore(model-profiles): refresh model profile data (#40994)
+chore(model-profiles): refresh model profile data (#40974)
+chore(deps): bump urllib3 from 2.7.0 to 2.8.0 in /libs/partners/openrouter (#40949)
+chore(model-profiles): refresh model profile data (#40924)
+chore(model-profiles): refresh model profile data (#40902)
+chore(model-profiles): refresh model profile data (#40869)
+chore(model-profiles): refresh model profile data (#40833)
+chore(model-profiles): refresh model profile data (#40804)
+chore(model-profiles): refresh model profile data (#40780)
+chore(model-profiles): refresh model profile data (#40747)
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.5
triage reasoningVersion 0.3.0 removes Python 3.10 support; teams on that version must stay on 0.2.x or upgrade.
signalsPython 3.10 support dropped · minimum version requirement increased to Python 3.11 · minimum langchain-core version bumped to 1.8.3 · breaking change for Python 3.10 users
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)
WORTH KNOWING API LangChain releases · 2026-11-18 20:35 UTC

langchain-openai 1.8.2 fixes async HTTP client sharing between models

Do:Upgrade to langchain-openai 1.8.2 if you use multiple OpenAI models concurrently in async code.
WHAT CHANGEDlangchain-openai 1.8.2 no longer shares the async HTTP client between models, adds support for GPT-6.2 Astra and Sol models, and sends langsmith_gateway headers from the client.
WHY IT MATTERSShared async HTTP clients between models can cause concurrency issues in applications using multiple OpenAI models simultaneously. The fix isolates client state per model instance.
WHAT TO DOUpgrade to langchain-openai 1.8.2 if you use multiple OpenAI models concurrently in async code. Test that existing async patterns still work as expected after the client isolation change.
Model-reported confidence 75% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## langchain-openai==1.8.2
+https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.8.2
+2026-11-18T16:53:07Z
+
+Changes since langchain-openai==1.8.1
+fix(openai): don't share the async HTTP client between models (#41944)
+feat(openai): support GPT-6.2 Astra and Sol (#41922)
+feat(openai): send langsmith_gateway headers from the client (#41901)
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.72, needed 0.5
triage reasoningNew model support and async client isolation are integration-relevant; header change affects observability pipeline.
signalsnew model support added (GPT-6.2 Astra and Sol) · async HTTP client behavior change (potential breaking change for concurrent usage) · new header transmission (langsmith_gateway) — integration/telemetry change
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)
ACTION NEEDED API LangChain releases · 2026-11-18 20:35 UTC

langchain-anthropic 1.7.11 drops Python 3.10 support

Do:Check your Python version with `python --version`.
WHAT CHANGEDlangchain-anthropic 1.7.11 removed Python 3.10 from supported versions. The release adds support for Claude Fable and Mythos 5.2 models, introduces langsmith_gateway header forwarding from the client, and changes ModelRefusalError behavior to raise for refusals when tool calls are not forced.
WHY IT MATTERSCode running on Python 3.10 will fail dependency resolution or runtime checks when this version is installed. Projects pinned to older langchain-anthropic versions cannot use the newly supported Claude models.
WHAT TO DOCheck your Python version with `python --version`. If 3.10, upgrade to 3.11+ before updating langchain-anthropic, or pin langchain-anthropic<1.7.11 in requirements.
Model-reported confidence 85% (not measured accuracy) Primary source ↗ written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## langchain-anthropic==1.7.11
+https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.7.11
+2026-11-18T15:20:21Z
+
+Changes since langchain-anthropic==1.7.10
+release(anthropic): 1.7.11 (#41929)
+feat(anthropic): send langsmith_gateway headers from the client (#41902)
+feat(anthropic): support Claude Fable and Mythos 5.2 (#41770)
+fix(anthropic): emit a chunk when an empty thinking block starts (#41907)
+feat(partners): Remove EOL'd Python 3.10 from partner packages (#41268)
+chore(model-profiles): refresh model profile data (#41791)
+chore(model-profiles): refresh model profile data (#41781)
+fix(anthropic): raise ModelRefusalError for refusals when tool calls are not forced (#41715)
+fix(anthropic): preserve nested caller on server calls and results (#41678)
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.72, needed 0.5
triage reasoningNew model support, dropped Python version, and authentication feature addition affect deployment and integration decisions.
signalsnew models added: Claude Fable and Mythos 5.2 · Python 3.10 support removed from partner packages · new authentication/integration feature: langsmith_gateway headers · error handling change: ModelRefusalError behavior with tool calls
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modelanthropic/claude-sonnet-4.5 ($0.00000)