Students pick a base model by leaderboard rank and then discover the license forbids commercial use, or the model has no quantized GGUF build, or the community support is thin. Here's the checklist that actually matters.
- License: permissive enough for what you plan to do with the result, including a resume portfolio piece or a startup.
- Size vs your hardware: can you fine-tune it with QLoRA on the VRAM you actually have, not the VRAM a benchmark assumed.
- Instruction-tuned base, not raw pretrained, unless your task genuinely needs to start from scratch.
- Community tooling: an active ecosystem means faster answers when something breaks mid-project.
- Context length: matches what your actual inputs need, not the largest number on the spec sheet.
Leaderboard rank predicts general capability, not whether a model is fine-tunable on your setup. See fine-tuning on a consumer GPU: the VRAM budget and small language models: 1B–3B use cases.
— Pranjul Rathour, GenAI Engineer from Kanpur, India. Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at any campus: pranjulrathour41@gmail.com.