Skip to content

fix: handle stale low context metadata for curated models#25464

Open
Jilazem wants to merge 1 commit into
NousResearch:mainfrom
Jilazem:fix/nemotron-context-underreport
Open

fix: handle stale low context metadata for curated models#25464
Jilazem wants to merge 1 commit into
NousResearch:mainfrom
Jilazem:fix/nemotron-context-underreport

Conversation

@Jilazem

@Jilazem Jilazem commented May 14, 2026

Copy link
Copy Markdown

Summary

Test Plan

  • python -m pytest tests/agent/test_model_metadata.py tests/run_agent/test_compression_feasibility.py -q

Fixes #24140

- Treat sub-64K provider/cache metadata as stale when a curated fallback is available
- Reuse longest-first fallback matching for metadata guards
- Add Nemotron-3-120B regression coverage

Fixes NousResearch#24140
@alt-glitch alt-glitch added type/bug Something isn't working P2 Medium — degraded but workaround exists comp/agent Core agent loop, run_agent.py, prompt builder labels May 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

comp/agent Core agent loop, run_agent.py, prompt builder P2 Medium — degraded but workaround exists type/bug Something isn't working

Projects

None yet

Development

Successfully merging this pull request may close these issues.

All models rejected with "context window below minimum 64,000 tokens" — Telegram completely down

2 participants