legion-ref/mesh-llm
1
0
Fork 0
mirror of https://github.com/Mesh-LLM/mesh-llm.git synced 2026-08-08 22:23:19 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(catalog): add Qwen3.6 27B and 35B-A3B to the model catalog (#1184)

Browse source 
Qwen3.6 is available on Hugging Face but absent from the bundled catalog, so
these models could only be run by typing a full repo reference. Add the dense
27B and the 35B-A3B MoE so they show up in model listings, search, and the
console catalog alongside Qwen3.5.

Both entries include the mmproj asset, since both models are vision-capable
(their configs carry vision_config and *ForConditionalGeneration
architectures), which makes the catalog advertise vision support for them.

No draft model is set: Qwen3.6 uses a 248320-token vocabulary, so the
Qwen3-0.6B draft used by the Qwen3.5 entry is not tokenizer-compatible.

Verified both download URLs return HTTP 200 (16.4GB and 20.8GB) and both
models serve and generate locally on Metal.

Co-authored-by: Michael Neale <14976+michaelneale@users.noreply.github.com>
This commit is contained in:
Michael Neale 2026-08-06 10:04:52 +10:00 • committed by GitHub
parent 4d400b338c
commit 32eedbe452
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 52 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

26
crates/mesh-client/src/models/catalog.json
View file

@ -255,6 +255,32 @@
"url": "https://huggingface.co/unsloth/Qwen3.5-27B-GGUF/resolve/main/mmproj-BF16.gguf"
}
},
{
"name": "Qwen3.6-27B-UD-Q4_K_XL",
"file": "Qwen3.6-27B-UD-Q4_K_XL.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-27B-GGUF/resolve/main/Qwen3.6-27B-UD-Q4_K_XL.gguf",
"size": "16.4GB",
"description": "Qwen3.6 27B dense, vision + text, strong reasoning and coding",
"draft": null,
"extra_files": [],
"mmproj": {
"file": "Qwen3.6-27B-mmproj-BF16.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-27B-GGUF/resolve/main/mmproj-BF16.gguf"
}
},
{
"name": "Qwen3.6-35B-A3B-UD-Q4_K_XL",
"file": "Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/resolve/main/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf",
"size": "20.8GB",
"description": "Qwen3.6 35B MoE (3B active), vision + text, fast reasoning and coding",
"draft": null,
"extra_files": [],
"mmproj": {
"file": "Qwen3.6-35B-A3B-mmproj-BF16.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/resolve/main/mmproj-BF16.gguf"
}
},
{
"name": "Qwen3-Coder-Next-Q4_K_M",
"file": "Qwen3-Coder-Next-Q4_K_M-00001-of-00004.gguf",

26
crates/mesh-llm-node/src/catalog.json
View file

@ -255,6 +255,32 @@
"url": "https://huggingface.co/unsloth/Qwen3.5-27B-GGUF/resolve/main/mmproj-BF16.gguf"
}
},
{
"name": "Qwen3.6-27B-UD-Q4_K_XL",
"file": "Qwen3.6-27B-UD-Q4_K_XL.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-27B-GGUF/resolve/main/Qwen3.6-27B-UD-Q4_K_XL.gguf",
"size": "16.4GB",
"description": "Qwen3.6 27B dense, vision + text, strong reasoning and coding",
"draft": null,
"extra_files": [],
"mmproj": {
"file": "Qwen3.6-27B-mmproj-BF16.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-27B-GGUF/resolve/main/mmproj-BF16.gguf"
}
},
{
"name": "Qwen3.6-35B-A3B-UD-Q4_K_XL",
"file": "Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/resolve/main/Qwen3.6-35B-A3B-UD-Q4_K_XL.gguf",
"size": "20.8GB",
"description": "Qwen3.6 35B MoE (3B active), vision + text, fast reasoning and coding",
"draft": null,
"extra_files": [],
"mmproj": {
"file": "Qwen3.6-35B-A3B-mmproj-BF16.gguf",
"url": "https://huggingface.co/unsloth/Qwen3.6-35B-A3B-GGUF/resolve/main/mmproj-BF16.gguf"
}
},
{
"name": "Qwen3-Coder-Next-Q4_K_M",
"file": "Qwen3-Coder-Next-Q4_K_M-00001-of-00004.gguf",