JM

akierum

·

AI & ML interests

None yet

Recent Activity

new activity 7 days ago

unsloth/Qwen3.8-27B-GGUF:Intel Arc B70 results: 32 T/s with UD-Q6_K, SYCL, MTP

new activity 10 days ago

bytkim/Qwen3.8-27B-pi-FP8:Works great :)

new activity 10 days ago

unsloth/Qwen3.8-27B-GGUF:Uniform GGUF quants silently break Qwen3.8-27B's deep thinking — reproduced on llama.cpp AND vLLM (short tasks unaffected)

View all activity

Organizations

None yet

New activity in unsloth/Qwen3.8-27B-GGUF 7 days ago

Intel Arc B70 results: 32 T/s with UD-Q6_K, SYCL, MTP

#114 opened 2 months ago by

New activity in bytkim/Qwen3.8-27B-pi-FP8 10 days ago

Works great :)

#1 opened 12 days ago by

New activity in unsloth/Qwen3.8-27B-GGUF 10 days ago

Uniform GGUF quants silently break Qwen3.8-27B's deep thinking — reproduced on llama.cpp AND vLLM (short tasks unaffected)

#135 opened 20 days ago by

New activity in peculiar-ragdoll/Dirk-Qwen3.8-27B-GGUF 10 days ago

Q8-XL version same bug as Unsloth - general.file_type Q4_K_M

#20 opened 10 days ago by

New activity in bytkim/Qwen3.8-27B-pi-GGUF 10 days ago

Please release XL quant

#5 opened 15 days ago by

New activity in unsloth/Qwen3.8-27B-GGUF 10 days ago

GGUF versions available via Torrent

#123 opened about 2 months ago by

Introducing Unsloth Dynamic v3 Qwen3.8

#74 opened 2 months ago by

New activity in ukisai/Swift-Qwen3.8-27B-GGUF about 1 month ago

Please add Q8-XL version to make replacement for Unsloth Xl version.

#6 opened about 1 month ago by

New activity in unsloth/Qwen3.8-Flash-Next-GGUF about 2 months ago

Run 170 tokens/s Qwen3.8-Flash with MTP! ⚡

#56 opened about 2 months ago by

replied to bytkim's post about 2 months ago

Thanks friend, does that mean there will be reasoning an non reasoning versions?

replied to bytkim's post about 2 months ago

Your model is best in class.

Waiting for Qwen 3.8 27B pi thinking model too.
Your models are the only ones passing that codeneedle test no making empty space errors on first row.

I would suggest releasing the XL versions too as BF16 is out of reach for everyone, and keep up the good work.

New activity in peculiar-ragdoll/Dirk-Qwen3.8-27B-GGUF about 2 months ago

Where is the normal Q8 version? to fit into 48GB VRAM with 256k context?

#13 opened about 2 months ago by

New activity in zai-org/GLM-5.3-Flash about 2 months ago

WE NEED 35B A3B

#3 opened 2 months ago by

New activity in bytkim/Qwen3.6-27B-MTP-pi-reasoning-GGUF 2 months ago

will there be qwen 3.8 27B version ?

#5 opened 2 months ago by

New activity in beyoru/KAT-Coder-V2.5-Dev-VL 2 months ago

the bartowski quantization works fine with mmproj from qwen no need for modifications

#1 opened 2 months ago by

New activity in Kwaipilot/KAT-Coder-V2.5-Dev 2 months ago

WOW! Not kidding, this model nuts! 💥

#29 opened 2 months ago by

New activity in z-lab/Qwen3.8-27B-DFlash2 2 months ago

for 3090 MTP and gguf is best for performance 250k context and accuracy.

#8 opened 2 months ago by

reacted to DavidAU's post with 😎🚀🧠 2 months ago

Post

22,831

Qwen3.8 - 27B - COLD FUSION

Reduction in thinking tokens to 1/10 to 1/2 "normal Qwen" across all three modes of thinking without loss of detail in thinking or output.

Increase in general intelligence too (all 7 critical benches); bench marks posted.

Trained using the COLD FUSION method (GAIN + UNSLOTH).
NOTE: This is NOT a heretic / uncensored version.

MTP and GGUF NEO MAX quants:
DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1-NM-DAU-NEO-MAX-MTP-GGUF

SOURCE:
DavidAU/Qwen3.8-27B-Cold-Fusion-GAIN-V1.1

7 replies

·