Hugging Face
Models
Datasets
Spaces
Buckets
new
Docs
Enterprise
Pricing
Website
Tasks
HuggingChat
Collections
Languages
Organizations
Community
Blog
Posts
Daily Papers
Hardware
Learn
Discord
Forum
GitHub
Solutions
Team & Enterprise
Hugging Face PRO
Enterprise Support
Inference Providers
Inference Endpoints
Storage Buckets
Log In
Sign Up
36.9
TFLOPS
ManniX
PRO
ManniX-ITA
38
6
28
Follow
21world's profile picture
3hP4's profile picture
Axway's profile picture
103 followers
ยท
22 following
https://github.com/mann1x
mann1x
AI & ML interests
None yet
Recent Activity
new
activity
about 6 hours ago
ManniX-ITA/Qwen3.6-27B-A3B-CoderX:
Best performing model of qwen3.6 family
posted
an
update
1 day ago
๐ JackOD-9B-Coder โ a 9B merge built to FINISH agentic coding tasks. Four-way omnimerge_v2 over Qwen3.5-9B: Jack = Qwopus3.5-9B-Coder (0.30), O = Ornith-1.5-9B (0.15), D = DeltaCoder (0.55). MTP head kept. ๐ Q6_K + imatrix, llama.cpp, greedy, lcb_v6_55 โ merge / base / DeltaCoder / Qwopus / Ornith: โก LiveCodeBench v6 (55 hard) โ 0.7818 / 0.7273 / 0.6364 / 0.6000 / 0.5818 โ HumanEval โ 0.8841 / 0.8902 / 0.9146 / 0.8537 / 0.7805 โ HumanEval+ โ 0.8232 / 0.8049 / 0.8232 / 0.7988 / 0.7073 ๐ค MultiPL-E โ 0.8033 / 0.8200 / 0.8000 / 0.8200 / 0.7267 ๐ IFEval โ 0.9100 / 0.9300 / 0.9200 / 0.8800 / 0.8200 ๐ฏ LCB beats every source AND the base: +5.45pp over the base, +14.54pp over DeltaCoder, its heaviest. ๐ And why. Same 55 problems, same cap, generations that NEVER terminated: DeltaCoder 25/55 ยท base 18/55 ยท Qwopus 8/55 ยท Ornith 2/55 ยท JackOD 1/55. That split is the thesis: DeltaCoder is the cohort's best coder and worst at stopping, Ornith the weakest and best at stopping. The merge takes BOTH. ๐ค tool-eval-bench hardmode, 5 seeds: JackOD 144.4 ยฑ4.7, second behind Ornith 145.6 ยฑ4.2, above base 142.0 โ all CIs overlap. But Autonomous Planning: JackOD 5.2/6, best of five, Ornith WORST at 2.8/6. Ornith stops reliably but plans worst โ it stops too early. The merge does both. ๐ง Serving: temp 0.6 / top_p 0.95 / top_k 20 + presence_penalty 1.5 โ the penalty stops it re-treading a tool call. Tool calling on llama.cpp needs --jinja. ๐งช Initial impression, limited testing: fixed a cline-harness task in 535s; A3B models want 1.5-4h at <50% success. ๐ฆ 25 GGUF tiers, every K/IQ imatrix-built incl Q6_K, plus a ContribDynamic ladder (per-tensor maps from our imatrix, Unsloth-UD style). ๐ https://huggingface.co/danielcherubini/Qwen3.5-DeltaCoder-9B ยท https://huggingface.co/ornith-ai/Ornith-1.5-9B ๐ https://huggingface.co/ManniX-ITA/JackOD-9B-Coder ๐ https://huggingface.co/ManniX-ITA/JackOD-9B-Coder-MTP-GGUF ๐ https://ollama.com/mannix/JackOD-9B-Coder
published
a model
1 day ago
ManniX-ITA/JackOD-9B-Coder-MTP-GGUF
View all activity
Organizations
None yet
ManniX-ITA
's datasets
1
Sort:ย Recently updated
ManniX-ITA/osync-code
Viewer
โข
Updated
Jan 12
โข
1
โข
26