Models
Datasets
Spaces
Buckets
Docs
Enterprise
Pricing
- Website
- Community
- Solutions
Log In
Sign Up

Collections

Discover the best community collections!

Collections trending this week

LFM2.5-2.6B-heretic — OpenVINO IR

2.6B dense abliterated LFM2.5 at int8 and int4, the small-model reference build.

blaj/LFM2.5-2.6B-heretic-int8-ov

Text Generation • Updated 19 days ago • 54
blaj/LFM2.5-2.6B-heretic-int4-ov

Text Generation • Updated 19 days ago • 55

Qwen2.5-VL-7B-Instruct-abliterated — OpenVINO IR

Abliterated Qwen2.5-VL 7B at int8 and int4 for local vision-language inference.

blaj/Qwen2.5-VL-7B-Instruct-abliterated-int8-ov

Image-Text-to-Text • Updated 19 days ago • 34
blaj/Qwen2.5-VL-7B-Instruct-abliterated-int4-ov

Image-Text-to-Text • Updated 19 days ago • 38

Qwen3.8-3.6-27B-blend — OpenVINO IR

The 3.6/27B blend at fp16, int8 and int4. Exposes hidden-state locators; no DFlash draft is published yet.

blaj/Qwen3.8-3.6-27B-blend-fp16-ov

Image-Text-to-Text • Updated 19 days ago • 38
blaj/Qwen3.8-3.6-27B-blend-int8-ov

Image-Text-to-Text • Updated 19 days ago • 42
blaj/Qwen3.8-3.6-27B-blend-int4-ov

Image-Text-to-Text • Updated 19 days ago • 40

Qwen3-4B — OpenVINO IR + DFlash

Qwen3-4B Base at int8/int4 and instruct at int4, with int8/int4 DFlash drafts that read five target layers.

blaj/Qwen3-4B-Base-int8-ov

Text Generation • Updated 19 days ago • 41
blaj/Qwen3-4B-Base-int4-ov

Text Generation • Updated 19 days ago • 41
blaj/Qwen3-4B-instruct-int4-ov

Text Generation • Updated 19 days ago • 32
blaj/Qwen3-4B-DFlash-int8-ov

Updated 19 days ago

Qwen3.8-27B-heretic-ara — OpenVINO IR + DFlash

64-layer Qwen3.8-27B abliterated target at int8/int4, plus matching DFlash drafts at fp16, int8 and int4.

blaj/Qwen3.8-27B-heretic-ara-int8-ov

Image-Text-to-Text • Updated 19 days ago • 38
blaj/Qwen3.8-27B-heretic-ara-int4-ov

Image-Text-to-Text • Updated 19 days ago • 59
blaj/Qwen3.8-27B-heretic-ara-DFlash-fp16-ov

Updated 19 days ago • 59
blaj/Qwen3.8-27B-heretic-ara-DFlash-int8-ov

Updated 19 days ago • 51

LFM2.5-8B-A1B-heretic — OpenVINO IR

8B A1B mixture-of-experts abliterated LFM2.5 at int8 and int4.

blaj/LFM2.5-8B-A1B-heretic-int8-ov

Text Generation • Updated 19 days ago • 46
blaj/LFM2.5-8B-A1B-heretic-int4-ov

Text Generation • Updated 19 days ago • 67

Llama-3.1-8B-Lexi-Uncensored-V2 — OpenVINO IR

Uncensored Llama-3.1-8B at fp16, int8 and int4. No hidden-state locators, so no DFlash pairing.

blaj/Llama-3.1-8B-Lexi-Uncensored-V2-fp16-ov

Text Generation • Updated 19 days ago • 303
blaj/Llama-3.1-8B-Lexi-Uncensored-V2-int8-ov

Text Generation • Updated 19 days ago • 30
blaj/Llama-3.1-8B-Lexi-Uncensored-V2-int4-ov

Text Generation • Updated 19 days ago • 47

Ornith-1.5-9B — OpenVINO IR + DFlash

Ornith-tuned 9B target at int8/int4 with the int4 DFlash draft that reads eight target layers.

blaj/Ornith-1.5-9B-int8-ov

Updated 19 days ago • 30
blaj/Ornith-1.5-9B-int4-ov

Updated 19 days ago • 82
blaj/Ornith-1.5-9B-DFlash-int4-ov

Updated 19 days ago • 44

Qwen3-8B — OpenVINO IR + DFlash

Qwen3-8B Base target at int4 plus the int8/int4 DFlash drafts that pair with it for speculative decoding.

blaj/Qwen3-8B-Base-int4-ov

Text Generation • Updated 19 days ago • 169 • 1
blaj/Qwen3-8B-DFlash-int8-ov

Updated 19 days ago
blaj/Qwen3-8B-DFlash-int4-ov

Updated 19 days ago

OpenVINO IR builds — by model family

11 family collections: every published precision of each target, plus the DFlash drafts that pair with it.

Qwen3.8-27B-heretic-ara — OpenVINO IR + DFlash

Collection

64-layer Qwen3.8-27B abliterated target at int8/int4, plus matching DFlash drafts at fp16, int8 and int4. • 5 items • Updated 19 days ago
Qwen3.5-9B-abliterated — OpenVINO IR + DFlash

Collection

32-layer abliterated Qwen3.5-9B target at int8/int4 with the matching DFlash drafts at bf16, int8 and int4. • 5 items • Updated 19 days ago
Qwen3-8B — OpenVINO IR + DFlash

Collection

Qwen3-8B Base target at int4 plus the int8/int4 DFlash drafts that pair with it for speculative decoding. • 3 items • Updated 19 days ago
Qwen3-4B — OpenVINO IR + DFlash

Collection

Qwen3-4B Base at int8/int4 and instruct at int4, with int8/int4 DFlash drafts that read five target layers. • 5 items • Updated 19 days ago

LFM2.5-2.6B-heretic — OpenVINO IR

2.6B dense abliterated LFM2.5 at int8 and int4, the small-model reference build.

blaj/LFM2.5-2.6B-heretic-int8-ov

Text Generation • Updated 19 days ago • 54
blaj/LFM2.5-2.6B-heretic-int4-ov

Text Generation • Updated 19 days ago • 55

LFM2.5-8B-A1B-heretic — OpenVINO IR

8B A1B mixture-of-experts abliterated LFM2.5 at int8 and int4.

blaj/LFM2.5-8B-A1B-heretic-int8-ov

Text Generation • Updated 19 days ago • 46
blaj/LFM2.5-8B-A1B-heretic-int4-ov

Text Generation • Updated 19 days ago • 67

Qwen2.5-VL-7B-Instruct-abliterated — OpenVINO IR

Abliterated Qwen2.5-VL 7B at int8 and int4 for local vision-language inference.

blaj/Qwen2.5-VL-7B-Instruct-abliterated-int8-ov

Image-Text-to-Text • Updated 19 days ago • 34
blaj/Qwen2.5-VL-7B-Instruct-abliterated-int4-ov

Image-Text-to-Text • Updated 19 days ago • 38

Llama-3.1-8B-Lexi-Uncensored-V2 — OpenVINO IR

Uncensored Llama-3.1-8B at fp16, int8 and int4. No hidden-state locators, so no DFlash pairing.

blaj/Llama-3.1-8B-Lexi-Uncensored-V2-fp16-ov

Text Generation • Updated 19 days ago • 303
blaj/Llama-3.1-8B-Lexi-Uncensored-V2-int8-ov

Text Generation • Updated 19 days ago • 30
blaj/Llama-3.1-8B-Lexi-Uncensored-V2-int4-ov

Text Generation • Updated 19 days ago • 47

Qwen3.8-3.6-27B-blend — OpenVINO IR

The 3.6/27B blend at fp16, int8 and int4. Exposes hidden-state locators; no DFlash draft is published yet.

blaj/Qwen3.8-3.6-27B-blend-fp16-ov

Image-Text-to-Text • Updated 19 days ago • 38
blaj/Qwen3.8-3.6-27B-blend-int8-ov

Image-Text-to-Text • Updated 19 days ago • 42
blaj/Qwen3.8-3.6-27B-blend-int4-ov

Image-Text-to-Text • Updated 19 days ago • 40

Ornith-1.5-9B — OpenVINO IR + DFlash

Ornith-tuned 9B target at int8/int4 with the int4 DFlash draft that reads eight target layers.

blaj/Ornith-1.5-9B-int8-ov

Updated 19 days ago • 30
blaj/Ornith-1.5-9B-int4-ov

Updated 19 days ago • 82
blaj/Ornith-1.5-9B-DFlash-int4-ov

Updated 19 days ago • 44

Qwen3-4B — OpenVINO IR + DFlash

Qwen3-4B Base at int8/int4 and instruct at int4, with int8/int4 DFlash drafts that read five target layers.

blaj/Qwen3-4B-Base-int8-ov

Text Generation • Updated 19 days ago • 41
blaj/Qwen3-4B-Base-int4-ov

Text Generation • Updated 19 days ago • 41
blaj/Qwen3-4B-instruct-int4-ov

Text Generation • Updated 19 days ago • 32
blaj/Qwen3-4B-DFlash-int8-ov

Updated 19 days ago

Qwen3-8B — OpenVINO IR + DFlash

Qwen3-8B Base target at int4 plus the int8/int4 DFlash drafts that pair with it for speculative decoding.

blaj/Qwen3-8B-Base-int4-ov

Text Generation • Updated 19 days ago • 169 • 1
blaj/Qwen3-8B-DFlash-int8-ov

Updated 19 days ago
blaj/Qwen3-8B-DFlash-int4-ov

Updated 19 days ago

Qwen3.8-27B-heretic-ara — OpenVINO IR + DFlash

64-layer Qwen3.8-27B abliterated target at int8/int4, plus matching DFlash drafts at fp16, int8 and int4.

blaj/Qwen3.8-27B-heretic-ara-int8-ov

Image-Text-to-Text • Updated 19 days ago • 38
blaj/Qwen3.8-27B-heretic-ara-int4-ov

Image-Text-to-Text • Updated 19 days ago • 59
blaj/Qwen3.8-27B-heretic-ara-DFlash-fp16-ov

Updated 19 days ago • 59
blaj/Qwen3.8-27B-heretic-ara-DFlash-int8-ov

Updated 19 days ago • 51

OpenVINO IR builds — by model family

11 family collections: every published precision of each target, plus the DFlash drafts that pair with it.

Qwen3.8-27B-heretic-ara — OpenVINO IR + DFlash

Collection

64-layer Qwen3.8-27B abliterated target at int8/int4, plus matching DFlash drafts at fp16, int8 and int4. • 5 items • Updated 19 days ago
Qwen3.5-9B-abliterated — OpenVINO IR + DFlash

Collection

32-layer abliterated Qwen3.5-9B target at int8/int4 with the matching DFlash drafts at bf16, int8 and int4. • 5 items • Updated 19 days ago
Qwen3-8B — OpenVINO IR + DFlash

Collection

Qwen3-8B Base target at int4 plus the int8/int4 DFlash drafts that pair with it for speculative decoding. • 3 items • Updated 19 days ago
Qwen3-4B — OpenVINO IR + DFlash

Collection

Qwen3-4B Base at int8/int4 and instruct at int4, with int8/int4 DFlash drafts that read five target layers. • 5 items • Updated 19 days ago

Previous
1
...
616
617
618
619
620
...
24,648
Next

Company

TOS Privacy About Careers

Website

Models Datasets Spaces Pricing Docs