作者： Jakub Rusinowski · 最后更新： 2026年6月15日
These are the strongest local models that fit entirely in 20 GB of VRAM, ranked by capability, with the quantization level and estimated tokens/sec needed to fit.
| Qwen 2.5 Family — Qwen 2.5 14B Instruct | Q4_K_M · 9.5 GB · ~84 tok/s on AMD Radeon RX 7900 XT |
| Qwen 3.5 (Legacy Listing — Unverified) — Qwen 3.5 122B-A10B (MoE) | Q4_K_M · 13.5 GB · ~400 tok/s on AMD Radeon RX 7900 XT |
| Gemma 4 (Legacy Listing — Unverified) — Gemma 4 27B ⭐ | Q4_K_M · 14 GB · ~57 tok/s on AMD Radeon RX 7900 XT |
| Qwen 3 — Qwen 3 14B | Q4_K_M · 9.5 GB · ~84 tok/s on AMD Radeon RX 7900 XT |
| Codestral — Codestral 22B | Q4_K_M · 13 GB · ~62 tok/s on AMD Radeon RX 7900 XT |
| DeepSeek R1 — DeepSeek R1 Distill Qwen 14B | Q4_K_M · 9.2 GB · ~87 tok/s on AMD Radeon RX 7900 XT |
| Phi-4 Family — Phi-4 (14B) | Q4_K_M · 9.2 GB · ~87 tok/s on AMD Radeon RX 7900 XT |
| Qwen3-Coder — Qwen3-Coder 80B-A3B (MoE) | Q4_K_M · 7.5 GB · ~400 tok/s on AMD Radeon RX 7900 XT |
| Gemma 3 — Gemma 3 27B Instruct | Q4_K_M · 16.5 GB · ~48 tok/s on AMD Radeon RX 7900 XT |
| Mistral Small 3.1 — Mistral Small 3.1 24B | Q4_K_M · 14.5 GB · ~55 tok/s on AMD Radeon RX 7900 XT |
| Qwen 3.6 — Qwen 3.6 27B | Q4_K_M · 16.8 GB · ~48 tok/s on AMD Radeon RX 7900 XT |
| InternLM 3 — InternLM 3 20B Instruct | Q4_K_M · 12.5 GB · ~64 tok/s on AMD Radeon RX 7900 XT |
| Mistral Family — Mistral Small 3 (24B) | Q4_K_M · 14.5 GB · ~55 tok/s on AMD Radeon RX 7900 XT |
| Qwen 3 — Qwen 3 30B-A3B (MoE) | Q4_K_M · 8 GB · ~400 tok/s on AMD Radeon RX 7900 XT |
| Granite 3.0 — Granite 3.0 8B Instruct | Q4_K_M · 5.5 GB · ~145 tok/s on AMD Radeon RX 7900 XT |
Qwen 2.5 Family, Qwen 3.5 (Legacy Listing — Unverified), Gemma 4 (Legacy Listing — Unverified), Qwen 3, Codestral all fit in 20 GB VRAM.
AMD Radeon RX 7900 XT, AMD Radeon RX 7900 XTX, Apple M2, Apple M3.