HARDWARE / VRAM

GeForce RTX 5090 · 32 GB

A fast single-GPU tier for 27B-class low-bit models, with limited remaining space for very long context or multiple concurrent users.

Source primaire

Consumer Blackwell GPU32 GB

01 / PRACTICAL CEILING

Nex-N2.5-mini

The highest listed result may be a tight or offloaded fit. For daily use, prefer the first result marked Confortable 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.

Ouvrir le guide

03 / MODEL ENVELOPE

Modèles open-weight représentatifs

Nex AGI

Nex-N2.5-mini

English

35BQ4 estimated budget

Mémoire estimée: ~21 GB
Contexte: 262,144 (config max_position_embeddings)

Alibaba Qwen

Qwen3.6 35B-A3B

35B3B active · Q4 / NVFP4 MoE

Mémoire estimée: ~25 GB
Contexte: 262K native

Alibaba Qwen

Qwen3 32B

32BGGUF Q4

Mémoire estimée: ~24.5 GB
Contexte: 32K+

Alibaba Qwen

Qwen3 30B-A3B

30B3B active · GGUF Q4 MoE

Mémoire estimée: ~22.5 GB
Contexte: 32K+

Alibaba Qwen

Qwen3.6 27B

27BQ4 / NVFP4 estimate

Mémoire estimée: ~18.5 GB
Contexte: 262K native

Alibaba Qwen

Qwen3.8 27B

27BQ4 estimate

Mémoire estimée: ~19.5 GB
Contexte: 262K native

Attention

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: Nex-N2.5-mini (Confortable), Qwen3.6 35B-A3B (Serré), Qwen3 32B (Serré), Qwen3 30B-A3B (Serré), Qwen3.6 27B (Confortable), Qwen3.8 27B (Confortable).