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.

Primaire bron

GB10 Grace Blackwell128 GB

01 / PRACTICAL CEILING

Qwen3.8 Flash Next

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

Open gids

03 / MODEL ENVELOPE

Representatieve open-weightmodellen

Alibaba Qwen

Qwen3.8 Flash Next

English

180BQ4 estimated budget

Geschat geheugen: ~108 GB
Context: 262,144 native; 1M extended

inclusionAI

Ling-3.0-flash

English

124BQ4 estimated budget

Geschat geheugen: ~74.4 GB
Context: Check selected runtime and checkpoint

OpenAI

gpt-oss-120b

117B5.1B active · MXFP4

Geschat geheugen: ~80 GB
Context: 128K

Llama 4 Scout

109B17B active · INT4 MoE

Geschat geheugen: ~78 GB
Context: 10M advertised

Llama 3.1 70B

70BGGUF Q4

Geschat geheugen: ~46 GB
Context: 128K

Nex AGI

Nex-N2.5-mini

English

35BQ4 estimated budget

Geschat geheugen: ~21 GB
Context: 262,144 (config max_position_embeddings)

Let op

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 (Krap), Ling-3.0-flash (Ruim), gpt-oss-120b (Ruim), Llama 4 Scout (Ruim), Llama 3.1 70B (Ruim), Nex-N2.5-mini (Ruim).