HARDWARE / VRAM

Workstation · 48 GB

70B Q4 serving begins here, with limited room for cache and concurrency.

RTX 6000 / dual GPU48 GB

01 / PRACTICAL CEILING

Qwen3 32B

The highest listed result may be a tight or offloaded fit. For daily use, prefer the first result marked Comfortable and keep context modest.

02 / DEFAULT RUNTIME

vLLM

Linux GPU servers, concurrency and OpenAI-compatible production APIs. CUDA, ROCm and XPU feature coverage differs. Driver, shared-memory and quantization compatibility are version-specific.

Abrir guia

03 / MODEL ENVELOPE

Modelos open-weight representativos

Alibaba Qwen

Qwen3 32B

32BGGUF Q4

Memória estimada: ~24.5 GB
Contexto: 32K+

Alibaba Qwen

Qwen3 30B-A3B

30B3B active · GGUF Q4 MoE

Memória estimada: ~22.5 GB
Contexto: 32K+

Google

Gemma 3 27B

27BINT4

Memória estimada: ~20.5 GB
Contexto: 128K

Mistral AI

Devstral Small 2 24B

24BGGUF Q4 / BF16

Memória estimada: ~18.5 GB
Contexto: 256K

OpenAI

gpt-oss-20b

21B3.6B active · MXFP4

Memória estimada: ~16 GB
Contexto: 128K

Alibaba Qwen

Qwen3 14B

14BGGUF Q4

Memória estimada: ~11.2 GB
Contexto: 32K+

Atenção

Fit is based on estimated total model memory. Driver support, KV cache, multimodal projectors, concurrency, and desktop applications can all reduce available headroom. Current fit: Qwen3 32B (Comfortable), Qwen3 30B-A3B (Comfortable), Gemma 3 27B (Comfortable), Devstral Small 2 24B (Comfortable), gpt-oss-20b (Comfortable), Qwen3 14B (Comfortable).