10 open-weight models fit in RTX 3060 12 GB at 4-bit, ranked by Analyzer Index.
| Model | Index | Memory | Est. tok/s | API $/month | Verdict |
|---|---|---|---|---|---|
| Ministral 3 3B 2512 Mistral3.8B | — | 2 GB | ~189 | $12.50 | pays back in 22 mo |
| Gemma 3 4B Google4.3B | — | 3 GB | ~167 | $5.25 | pays back in 53 mo |
| Stable Diffusion 3.5 Large Stability AI8.1B | — | 5 GB | ~88.9 | — | no API price |
| Qwen3 VL 8B Instruct Alibaba (Qwen)8.8B | — | 5 GB | ~81.8 | $43.75 | pays back in 6 mo |
| Ministral 3 8B 2512 Mistral8.9B | — | 5 GB | ~80.9 | $18.75 | pays back in 15 mo |
| Qwen3.5-9B Alibaba (Qwen)9.7B | — | 6 GB | ~74.2 | $11.25 | pays back in 25 mo |
| FLUX.1 [dev] Black Forest Labs11.9B | — | 7 GB | ~60.5 | — | no API price |
| Gemma 3 12B Google12.2B | — | 7 GB | ~59.0 | $7.00 | pays back in 40 mo |
| Ministral 3 14B 2512 Mistral13.9B | — | 8 GB | ~51.8 | $25.00 | pays back in 11 mo |
| Qwen3 14B Alibaba (Qwen)14.8B | — | 9 GB | ~48.6 | $15.50 | pays back in 18 mo |
Memory need is weights × 1.2 for runtime overhead. Speed is a theoretical single-stream ceiling (memory bandwidth ÷ active-weight bytes); real engines usually reach 50–80% of it, and batching many requests raises total throughput well beyond one stream. Rental and purchase prices are typical figures checked 2026-10-14; power and engineering time are not included.