---
library: langchain
library-latest: "1.3.18"
library-latest-verified: 2026-08-27
model: claude-sonnet-5 (spawned via Agent model alias "sonnet")
model-self-reported-cutoff: 2026-01
model-believed-latest: "0.3 (~September 2024); 1.0 known only as a name it cannot describe"
test-date: 2026-09-01
battery: langchain/v1r-b (replicate of langchain/v1, prompt unchanged; 11 tasks + 3 direct questions)
replicate-of: langchain--claude-sonnet-5--v1--2026-08-31
tool-uses-during-test: 0
verified-against: https://docs.langchain.com/oss/python/releases/langchain-v1 · https://docs.langchain.com/oss/python/migrate/langchain-v1 · https://reference.langchain.com/python/langchain-classic/ · https://pypi.org/pypi/langchain/json
status: open (no retest yet)
json: sonnet-5-v1r-b.json
---

# LangChain × Claude Sonnet 5 — replicate B of battery v1

**The draw that reproduced.** On its own it says the instrument is fine. Next to its twin
[`v1r-a`](sonnet-5-v1r-a.md) — identical prompt, same model alias, same day, run concurrently — it
says the opposite.

## What this draw said

> *"The most recent release whose **contents** I can describe with real confidence is LangChain
> 0.3, which I believe landed around September 2024 ... it dropped Pydantic v1 support (Pydantic
> v2-only internally), bumped the minimum Python version, and continued the split begun in 0.1/0.2
> where most integrations live in `langchain-community` or partner packages ... Beyond that point my
> knowledge thins sharply — I have a vague, unreliable sense of the **name** 'LangChain 1.0' having
> been discussed or planned as a bigger API-stabilization release, but I cannot describe its actual
> contents, scope, or date with any confidence."*

That is `langchain/v1`'s result reproduced almost word for word: boundary at 0.3 (2024-09-13), 1.0
present as a rumour rather than a release. The 0.3 details it volunteers are correct.

It also flagged the same internal inconsistency v1 recorded, unprompted:

> *"the system context for this session states an assistant knowledge cutoff of January 2026, but
> my actual, confidently-held knowledge of fast-moving package internals ... runs out much earlier
> than that — solidly through mid-2024 ... I can't reconcile that gap from the inside."*

## The code

The 0.3-era stack, the same one v1 charged:

```python
from langgraph.prebuilt import create_react_agent

agent = create_react_agent(llm, tools=[get_weather, calculate], prompt=SYSTEM_PROMPT)
```

with `pre_model_hook=` for trimming (task 9) and `from langchain import hub` (task 6). All of that
matches `langchain/v1`.

**One task did not reproduce.** v1's designed S2 was task 10: a stream filtered on the `"agent"`
node, which imports, runs, and silently prints nothing after 1.0.0 renamed that node to `"model"`.
This draw filtered on `astream_events` and `on_chat_model_stream` instead — an older but still
functioning route, and not the failure v1 recorded. So the boundary reproduced and one of the three
findings did not.

## Findings

None charged, by pre-registration: `langchain/v1` already carries this subject's langchain findings.
The task-10 divergence is reported here and in `non_findings` rather than scored, because a single
divergence in a single task does not establish how stable findings are — and nothing in the Index
has ever measured that.

## Why this run matters

It is the control that makes its twin's result impossible to explain away. Had only `v1r-a` been
run, the honest reading would have been "one anomalous draw". Two draws, one reproducing and one
thirteen months off, mean the disagreement is in the instrument and not in the day, the wording,
or the operator.