Codigger
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 1
- 128GB Unified-Memory Mini PCs Bring Large-Model Inference to the Desktop
掘金·人工智能 argues that 128GB unified-memory mini PCs are becoming a practical desktop platform for running 70B–100B-class models locally, with most of the shared CPU-GPU memory pool available for model weights and KV cache. It reports that Ollama defaults to a 4K context below 24GB and supports 256K at 48GB or more, while Qwen3.6 35B MoE reached 20 tokens per second on a 64GB laptop with integrated graphics, though community reports say the model can lose track of multi-step tasks. The article also compares Nvidia DGX Spark with an AMD mini PC and closes by promoting Codigger’s software-led approach to local inference, cloud workstations, and agent orchestration.
掘金 · 人工智能 · 🔥 0
Experience and discussion from the community
Share my Codigger experienceAsk about Codigger
Nobody has shared their experience with Codigger yet.