Duplicated from UnstableLlama/Jackrong_Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2-exl3-8.00bpw

UnstableLlama
/

Jackrong_Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2-exl3-3.00bpw

Model card Files Files and versions

>> Jackrong / Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2-exl3

[QUANTS BY :: UnstableLlama]

// Information

exl3 quantizations of Jackrong's Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2 via exllamav3.

2.10bpw, 3.00bpw, 4.00bpw, 5.00bpw & 8.00bpw up now.
more sizes in progress...

Repo generated automatically with ezexl3.

// Repo Data

Quantization graph

REVISION	GiB	KL DIV	PPL
2.00bpw	9.84	0.1810	7.9225
2.10bpw	10.09	0.1454	7.5339
3.00bpw	12.67	0.0399	6.9014
3.50bpw	13.99	0.0265	6.8728
4.00bpw	15.50	0.0141	6.7928
5.00bpw	18.34	0.0034	6.7620
6.00bpw	21.17	0.0039	6.7539
8.00bpw	26.83	0.0004	6.7557
bf16	51.75	0.0000	6.7555

SVG Catbench

2.00 bpw	2.10 bpw	3.00 bpw	3.50 bpw
4.00 bpw	5.00 bpw	6.00 bpw	8.00 bpw

// CLI Download

hf download UnstableLlama/Jackrong_Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2-exl3-4.00bpw --local-dir ./qwen3.5-27b-exl3-4.00bpw

Downloads last month: 16

Safetensors

Model size

7B params

Tensor type

F16

·

I16

·

F32

·

BF16

·

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for UnstableLlama/Jackrong_Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2-exl3-3.00bpw

Base model

Qwen/Qwen3.5-27B

Adapter

Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-v2

Quantized

(21)

this model