AWQ **W4A16** (4-bit weights, group size 128) quantization of `Qwen/Qwen3-Coder-30B-A3B-Instruct`, produced with llm-compressor for serving on a single 24 GB GPU with vLLM. Qwen3-Coder-30B-A3B is a Mixture-of-Experts model (128 experts, 8 active per token) using standard **GQA** attention — so it serves on consumer Ada GPUs with the stock attention backends, no special kernels required. Notes: - At `--max-model-len 65536` on a 24 GB card the KV pool holds ~118K tokens (~1.8x concurrency). The base model's native context is 262K; 64K is the practical cap at 24 GB. - Tool calling works via vLLM's first-class `qwen3_coder` parser. - Non-thinking model — no reasoning parser needed. - Pass `--override-generation-config` if your client doesn't send sampling params, or vLLM will default to `temperature=1.0`, which is far too hot for coding.

번역 대기 중 · 원문 표시

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗

DavidAU-Qwen3.8-27B-TURBO-Fable-Cold-Fusion-735-882-Heretic-Uncensored-NM-DAU-EXL3-4.00bpw-HB8

darkbit1001

基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。

기본 정보 등록 · 상세 정보 보완 예정

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗

XORTRON.CriminalComputing.LARGE.2026.3

darkc0de

基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。

기본 정보 등록 · 상세 정보 보완 예정

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗

LFM2.5-8B-A1B-Opus-Distil-Q5_K_M-imat-GGUF

DarkKitsune

基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。

기본 정보 등록 · 상세 정보 보완 예정

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗

Qwen3.6-Candid-Coder-35B-A3B

DarkKitsune

基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗

Qwen3.6-Qworus-35B-A3B-V1-APEX-GGUF

DarkKitsune

This is a DELLA merge of the following finetuned & post-trained models: | Model Repo | Purpose | | --- | --- | | EryriLabs/Ornith-1.5-35B-A3B-BigBang-MTP-GGUF | The base model, for terminal use, coding, software architecture, planning. Also provides MTP support | | Kwaipilot/KAT-Coder-V2.5-Dev | Contributes efficient coding with fewer tokens over more languages | | Accio-Lab/occamy-1.0 | Contributes general agent abilities, also a semi-official finetune | | AllSpark-Research/CompoWorld | Also contributes general agent abilities, different strengths/weaknesses from occamy | The result is an intelligent and efficient hybrid model for coding, terminal agent work, planning, designing, and conversation. Although most of this model's weights were merged with with a method that drops weights, more sensitive tensors such as the router gates and shared experts, were protected by using a density of 1.0 and epsilon of 0.0 to disable sparsification.

번역 대기 중 · 원문 표시

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗

Qwen3.8-Flash-Next-GSQ-RCO-IQ3_XXS-PLE-50pct

DarkKitsune

This model is ISTA-DASLab's GSQ-RCO IQ3_XXS quant of Qwen3.8-Flash-Next, but with the PLE (ngram) table pruned to 50% of its original parameters. This reduces the total size of the model further in return for an increase in perplexity. If using mmap without mlock, this probably won't make any noticable difference in VRAM usage, but with mmap *and* mlock (which boosts prefill speeds and lowers latency), there should be significantly less VRAM usage compared to the unpruned PLE table. The resulting model is ~61 GB (20% smaller than the original quant which is ~76GB), fitting nicely on 128 GB systems with ample room for the KV cache and other resources, even with mlock enabled. It is capable of chat, reasoning, basic tool calls & coding, but may stumble on more complex tasks. Compared to DarkKitsune/Qwen3.8-Flash-Next-UD-IQ3_XXS-PLE-25pct, this quant offers higher accuracy, and slightly higher performance.

번역 대기 중 · 원문 표시

텍스트 생성

입력 텍스트→ 출력 텍스트

요금 미확인멀티모달 및 채널 요금 별도 확인

자세히 보기 ↗

비교 추가출처 / 문서 ↗