01What it can do
- Apache 2.0 license with no restrictive clause
- Native vision: images, documents, and videos
- 262k native context
- Major improvement on agentic coding vs Qwen 3.6-27B
- Official Ollama tag + MLX builds for Apple Silicon
- —18 GB of model weights in Q4_K_M: 24 GB of VRAM to stay 100% on the GPU
- —Moderate local throughput (~9–22 tok/s) and a thinking mode that uses a lot of tokens
- —Manufacturer benchmarks, not independently reproduced
- —The 1M context requires vLLM/SGLang, inaccessible through Ollama
05Install
Install Ollama for your OS. Check the model and its quantization before downloading. Start with 4096 tokens of context, then check placement with ollama ps. A command below is not proof that a test was run on your machine.
02Required memory
Approximate GPU VRAM required to run this model, including 4k tokens of context overhead. For a longer context, add ~1 GB per 8k-token increment.
What hardware do you need for Qwen 3.8 27B?
To run Qwen 3.8 27B locally with Q4 quantization, you need about 16 GB of VRAM. An option to compare: GMKtec EVO-X2 64GB / 1TB (Ryzen AI Max+ 395) — leave some headroom for the system and context; check engine compatibility with the GPU.
Why this choice? Our complete guide on GMKtec EVO-X2 64GB / 1TB (Ryzen AI Max+ 395) →
Affiliate links — commission possible at no extra cost to you; independent recommendation. As an Amazon Associate, BestLLMfor earns from qualifying purchases.
On the go: Qwen 3.8 27B also runs on a RTX laptop PC (16 GB of VRAM) →
03Expected speed
Tokens generated per second in Q4_K_M, 4k context. Beyond 20 t/s, reading is comfortable. Below 10 t/s, that's just for testing.
04Public benchmarks
Scores reproduced from model cards or MMLU-Pro / community sources. Unit: % correct answers.