WhichLLM tool questioned after suggesting 27B models for 4GB VRAM
A Reddit user finds the model recommender lists incompatible LLMs for limited hardware.
Deep Dive
A developer with 4–6GB VRAM laptops used an online tool to find models their machine can run and was surprised to see it recommended gpt-oss-20b and qwen3.6-27b. They also noted that the tool’s RAM and free disk readings were incorrect, possibly due to Linux running inside WSL. The developer is not very knowledgeable about local models, having only used Ollama before.
Key Points
- WhichLLM recommended 20B and 27B parameter models for a laptop with only 4-6GB VRAM, suggesting poor accuracy.
- The tool misreported RAM and disk capacities, possibly due to WSL's virtualization layer distorting hardware detection.
- Developers using limited hardware (4-6GB VRAM) need reliable tools to find viable sub-7B parameter models like Qwen 2.5-Coder 3B.
Why It Matters
Accurate LLM compatibility tools are critical for developers on budget laptops, but WhichLLM's errors could lead to wasted time and false expectations.