Add RedHatAI/Qwen3.8-27B-INT4 model config - #54
Merged
taagarwa-rh merged 3 commits intoSep 21, 2026
Merged
Conversation
Register the INT4-quantized Qwen3.8-27B so the job queue can serve it for Michael's intake request. Mirrors the BF16/FP8 siblings (fp8 KV cache, 262144 context) so the leaderboard comparison isolates quantization impact.
Add ANTHROPIC/OPENAI/OPENROUTER key placeholders needed when submitting jobs with server_url=openrouter.
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Advanced Run ID: Comment |
Collaborator
|
LGTM! What concurrency did you get with this configuration? Please record it in our doc |
Collaborator
Author
|
Measured 16.49x on a single me west1 B200 (262144 context, fp8 KV cache), using the config in this PR. |
Collaborator
|
Ok great, I'll get this merged. Please record it in the doc linked above. In the future, please do not bump the version as part of a feature branch. That way if your feature branch is delayed, it doesn't accidentally revert the version |
taagarwa-rh
approved these changes
Sep 21, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds the INT4-quantized Qwen3.8-27B model config to support Michael's intake request (
rounakbende10/rh-swe-bench), to be served on Nebius B200.RedHatAI/Qwen3.8-27B-INT4ModelConfig, mirroring the BF16/FP8 siblings (fp8 KV cache, 262144 context) so the leaderboard comparison isolates quantization impactMODEL_CONFIGS+ test coverage.env.example0.2.4 → 0.2.5(rebuilds thecoding_agent_benchimage with these changes)