Catalog: add Muse Glimmer 30B, and a gptq-4bit board for Nemotron 3.5 Lightning - #31
Merged
jackwsmth merged 2 commits intoSep 18, 2026
Merged
Conversation
Three catalog additions, each tied to a measured result waiting to be submitted: - Muse Glimmer 30B (facebook/Muse-Glimmer-30B, dense 30B) with a q4_k_xl board, plus the matching QUANTS entry: Unsloth Dynamic Q4_K_XL is a distinct artifact class from Q4_K_M and gets its own board instead of being folded into one. - Nemotron 3.5 Lightning 30B A3B gains a gptq-4bit board for the GPTQ-INT4 G64 checkpoint its vLLM results run on, matching how the Qwen3.5 boards separate gptq-4bit from int4.
|
@SergiioB is attempting to deploy a commit to the Community Labs Team on Vercel. A member of the Team first needs to authorize it. |
No result files to import. |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What changes (one file, three additions)
muse-glimmer-30b): Meta's dense 30B, sourcefacebook/Muse-Glimmer-30B. Board:q4_k_xlonly — the one format actually run in the wild (Unsloth Dynamic GGUF), per the "list only the ones people actually run" rule. No logo: monogram tile is fine.q4_k_xl:GGUF, Unsloth Dynamic Q4_K_XL. Deliberately not folded intoq4_k_m: a dynamic Q4_K_XL artifact is a different weight layout than a stock Q4_K_M, and the boards are per-quantization.gptq-4biton its board — the GPTQ-INT4 G64 checkpoint its vLLM results run on has no board today.gptq-4bitalready exists inQUANTSand is already used this way by the Qwen3.5 boards (int4 = OpenVINO/IPEX weight-only, gptq-4bit = GPTQ), so this follows the established taxonomy instead of stretchingint4.Why these, and not more
autoroundtherefore also stays out ofQUANTSfor now: no live results need it.awqchanges — the id exists andqwen3-6-27balready carries anawq-4bitboard; nothing to add.Unblocks measured results
Once merged, I'll submit two held-back n=5 results that currently have no valid boards: Nemotron 3.5 Lightning DFlash n7 (186.61 tok/s, vLLM,
gptq-4bit) and Muse Glimmer 30B DFlash n2 (26.8 tok/s, llama.cpp,q4_k_xl). Both measured on one Arc Pro B70 with full coordinates.Verification
npm run catalog:validate: ✅ 46 hardware, 32 models, 15 quants, 7 runtimes.npm run typecheckandnpm run build: ✅.