Issue Station
Sign In
Tools hereToolbox

hermes - 💡(How to fix) Fix fix(metadata): qwen3.6-plus falls back to generic qwen 131K context instead of 1M [2 pull requests]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…

Root Cause

agent/model_metadata.py has qwen: 131072 as a catch-all, but qwen3.6-plus does not match this key. All new Qwen API models without explicit entries silently inherit wrong context limit.

Fix Action

Fixed

Code Example

qwen3.6-plus: 1048576,  # 1M context
RAW_BUFFERClick to expand / collapse

Describe the bug

qwen3.6-plus is missing from agent/model_metadata.py → DEFAULT_CONTEXT_LENGTHS. The model name does not match any specific entry, so it falls back to the generic qwen 131072 entry.

Actual context: 1,048,576 tokens (1M) Effective context in Hermes: 131,072 tokens (128K)

Impact

  • Compression threshold auto-lowered from 524,288 to 131,072 tokens
  • Unnecessary context compression triggers in long sessions
  • Degraded reasoning quality due to premature compression
  • Misleading warnings about context mismatch

Reproduction

Configure qwen3.6-plus as the main model via Alibaba/DashScope provider. Start a session, Hermes warns that compression model context 131K is lower than the threshold derived from main model.

Root cause

agent/model_metadata.py has qwen: 131072 as a catch-all, but qwen3.6-plus does not match this key. All new Qwen API models without explicit entries silently inherit wrong context limit.

Proposed fix

Add explicit entry in DEFAULT_CONTEXT_LENGTHS before the generic qwen fallback:

qwen3.6-plus: 1048576,  # 1M context

Affected versions

v2026.5.16 current release

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#training loop#device allocation#model download#tokenizer error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING