The Python AI stack after each model’s training cutoff · September 2026 · since-cutoff on GitHub

A coding model knows each library as it was when its training data ends. The libraries keep moving. To see how far, I pinned 36 widely used Python AI and LLM libraries at their releases of 26 September 2026 (the list) and ran since-cutoff scan (version 0.2.0) against the training cutoff of 21 models from 8 vendors: OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen, Moonshot and Mistral. For each library, the scan takes the release that was current at the model’s cutoff and diffs its public API against the pinned release, statically. No model was called for this page.

What it shows

• For every model, most of the stack changed its public API after the cutoff. Even the newest model here, Claude Opus 5.5 (training cutoff June 2026), predates a public API break in 20 of the 36 libraries, and 6 of them are on a new major version since: openai 2.44 → 3.19, anthropic 0.115 → 1.8, huggingface-hub 1.21 → 2.0, mcp 1.28 → 2.2, pinecone 9 → 10 and sentence-transformers 5 → 6.
• Models with early-2025 cutoffs have the most to catch up on. For Gemini 2.5 Pro and Claude Haiku 4.5 (cutoffs January and February 2025), 33 of the 36 libraries broke their public API since, and 21 are on a new major version, including openai, anthropic, langchain, langgraph, transformers, pandas and huggingface-hub.
• Older models count lower because much of the stack did not exist yet. GPT-4o’s training data (September 2023) predates any release of 13 of the 36 packages under their current names, among them langgraph, mcp, crewai, smolagents and google-genai. A model cannot write stale code for a library it has never seen, but it cannot write correct code for it either.

By model

model	vendor	training cutoff	new release since (of 36)	did not exist yet	new major version	public API breaks	new deprecations
gpt-4o	OpenAI	2023-09-30	36	13	15	21	6
gpt-4.1	OpenAI	2024-04-30	36	6	19	28	11
o3	OpenAI	2024-05-31	36	5	20	29	12
gpt-5	OpenAI	2024-09-30	36	4	18	30	12
kimi-k2	Moonshot	2024-10-31	36	4	18	30	12
mistral-large-latest	Mistral	2024-11-30	36	2	20	31	13
gemini-2.5-pro	Google	2025-01-31	35	0	21	33	13
claude-haiku-4-5	Anthropic	2025-02-28	35	0	21	33	13
claude-opus-4-6	Anthropic	2025-05-31	35	0	20	33	14
deepseek-v4-flash	DeepSeek	2025-05-31	35	0	20	33	14
claude-sonnet-4-5	Anthropic	2025-07-31	35	0	20	32	15
qwen3-coder	Qwen	2025-07-31	35	0	20	32	15
gpt-5.4	OpenAI	2025-08-31	35	0	19	32	14
gpt-5.5	OpenAI	2025-12-01	34	0	13	29	13
devstral-medium-latest	Mistral	2025-12-31	34	0	13	29	12
claude-opus-4-7	Anthropic	2026-01-31	34	0	11	29	12
grok-4.6	xAI	2026-02-01	34	0	11	29	12
gemini-3.7-flash	Google	2026-03-31	34	0	10	25	12
gpt-6-sol	OpenAI	2026-04-20	34	0	10	25	11
grok-4.7	xAI	2026-05-31	31	0	8	24	11
claude-opus-5-5	Anthropic	2026-06-30	31	0	6	20	9

“New release since” includes the libraries that did not exist yet. “Public API breaks” counts libraries with at least one public name or parameter removed, moved, newly required or changed in kind; “new deprecations” counts libraries that newly mark a public API as deprecated.

What this does and does not say

• It counts libraries, not mistakes. A library with a flagged break is one where code written from the model’s memory can fail; whether a model actually writes the old call is what since-cutoff run measures, per project. On one sample project, measured with since-cutoff 0.1.0, Claude Haiku 4.5 and Claude Opus 4.6 wrote stale code (valid for the version they learned, broken for the pinned one) for 3 of 5 and 2 of 4 probed libraries (results).
• It is a static diff. It sees names, parameters and deprecation markers in the public API, not behaviour changes behind an unchanged signature. Large libraries such as transformers report many changes, so this page counts libraries rather than changes.
• Cutoff dates come from models.dev. Vendors define them differently (Anthropic, for example, publishes both a “reliable knowledge” and a later “training data” cutoff), and models know the last months before their cutoff poorly, so the real gap can start earlier than the date suggests.
• “Did not exist yet” means no release under that PyPI name before the cutoff. Some projects existed under another name (dspy was dspy-ai).

Reproduce it

The pinned requirements file and the script that builds this page’s table and chart are in examples/ai-stack. No API key is needed: scan makes no model calls.

pip install since-cutoff
cd examples/ai-stack
mkdir -p scans
since-cutoff scan --model openai:gpt-5 --json > scans/gpt-5.json
since-cutoff scan --model anthropic:claude-opus-5-5 --json > scans/claude-opus-5-5.json
python stack_report.py scans      # writes stack.md, stack.json and the two chart SVGs

Your own project is the more useful test: uvx since-cutoff scan in its root shows what your model has not seen, and the MCP server lets your agent ask before it writes the code.