Open source #Open models#Coding agents Nov 7, 2026 · Originally published Nov 5

Foundry Local runs Qwen3.8-27B at 4-bit in about 18 GB, on 24 GB consumer GPUs with up to 64K context, as a GitHub Copilot model provider

Microsoft optimizes Qwen3.8-27B for local coding on NVIDIA GPUs and wires it into GitHub Copilot through BYOK, so everyday coding can stay on the machine.

Image: Microsoft Foundry Blog
Primary source Microsoft Foundry Blog · devblogs.microsoft.com Read the original ↗

// Key points

  • With 4-bit quantization the model is about 18 GB and runs on consumer GPUs with 24 GB of VRAM with context windows up to 64K tokens
  • The stack uses DFlash2 speculative decoding, INT4 weights, KV-cache quantization, CPU offloading, prefix caching and continuous batching
  • Against llama.cpp on an RTX 5090 and a DGX Spark, Foundry Local generated faster at every tested prompt length, with the gap widening beyond 131K tokens
  • Requires Foundry Local CLI 0.13.0 or later; Microsoft notes local models may reason less well than cloud models and suggests a hybrid workflow

Builder's takeA 27B model on a 24 GB card inside Copilot is the first local setup I'd actually hand to the team for routine edits and test writing. I'd route code that can't leave the office to it, keep cloud models for planning and hard bugs, and measure how often people have to switch back.

// Background · from #Open models

Full timeline →
  1. Nov 7 NVIDIA's TensorRT LLM VisualGen speeds FLUX 3 Image Klein 3.46x on one B200, cutting 2048x2048 text-to-image from 22.36 to 6.46 seconds
  2. Nov 5 Hugging Face proposes Model Integration Contracts for Transformers: tiny real-architecture fixtures on CPU that would have caught a UMT5 bug that sat in main for almost two months
  3. Nov 5 IBM Research's Palette builds slide decks with a 20B open-weight model and one LoRA adapter, scoring 59.7 on PresentBench versus 60.8 for NotebookLM

Every Monday, the week in AI in one email

Every Monday morning: the 6 AI stories that mattered most in the past seven days, with my take, in your inbox.

Your email is only used to send the weekly and is never shared. Every email has a one-click unsubscribe link. See the latest issue →