HARDWARE / UNIFIED MEMORY

NVIDIA DGX Spark · 128 GB

A compact local AI system with 128 GB coherent memory. It can hold large low-bit models, but decode speed and software support remain model-specific.

Sumber primer

GB10 Grace Blackwell128 GB

01 / PRACTICAL CEILING

Qwen3.8 Flash Next REAP-288

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.

Buka panduan

03 / MODEL ENVELOPE

Model open-weight perwakilan

Community REAP build

Qwen3.8 Flash Next REAP-288

176BREAP-288 · MLX 4-bit · PLE on NVMe

Estimasi memori: ~40.57 GB
Konteks: 7.6K recommended on measured 48 GB Mac

OpenAI

gpt-oss-120b

117B5.1B active · MXFP4

Estimasi memori: ~80 GB
Konteks: 128K

Llama 4 Scout

109B17B active · INT4 MoE

Estimasi memori: ~78 GB
Konteks: 10M advertised

Llama 3.1 70B

70BGGUF Q4

Estimasi memori: ~46 GB
Konteks: 128K

Alibaba Qwen

Qwen3.6 35B-A3B

35B3B active · Q4 / NVFP4 MoE

Estimasi memori: ~25 GB
Konteks: 262K native

Alibaba Qwen

Qwen3 32B

32BGGUF Q4

Estimasi memori: ~24.5 GB
Konteks: 32K+

Perhatian

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: Qwen3.8 Flash Next REAP-288 (Comfortable), gpt-oss-120b (Comfortable), Llama 4 Scout (Comfortable), Llama 3.1 70B (Comfortable), Qwen3.6 35B-A3B (Comfortable), Qwen3 32B (Comfortable).