Written by Jakub Rusinowski · Last updated July 12, 2026
Best all-round pick: Gemma 4 27B ⭐
The Apple M2 has 24 GB of unified memory, of which about 18 GB is available to a model. The largest model it can hold is Qwen 3.6 27B (28B, 17.6 GB at Q4_K_M).
| Model | Score | Memory at Q4_K_M | Context | Licence |
|---|---|---|---|---|
| Gemma 4 27B ⭐ | 97.4 | 17.1 GB | 125K | Gemma License (commercial OK) |
| Mistral Small 3.1 24B | 95.9 | 15 GB | 125K | Apache 2.0 |
| Gemma 3 27B Instruct | 93.8 | 17.1 GB | 128K | Gemma |
| Qwen 3.6 27B | 93.7 | 17.6 GB | 256K | Apache-2.0 |
| Qwen 3.5 14B | 93.2 | 9.3 GB | 125K | Apache 2.0 |
| Qwen 3 8B | 92.3 | 5.8 GB | 125K | Apache 2.0 |
| Workload | Recommended on this GPU |
|---|---|
| Coding | Gemma 4 27B ⭐ (96.8) Qwen 3.6 27B (96.8) Qwen 3 14B (95.2) |
| General assistant | Gemma 4 27B ⭐ (97.4) Mistral Small 3.1 24B (95.9) Gemma 3 27B Instruct (93.8) |
| Reasoning | Gemma 4 27B ⭐ (100) Qwen 3.6 27B (96.5) Qwen 3.5 27B (96) |
| RAG | Qwen 3.6 27B (96.7) Gemma 4 26B-A4B (94.1) Ternary Bonsai 27B (90.2) |
| Agents | Ternary Bonsai 27B (94) GLM-6 9B (84.3) Qwen 3.5 14B (83.8) |
| Vision | Gemma 4 27B ⭐ (96.8) Mistral Small 3.1 24B (95.2) Qwen 3.5 14B (94.7) |
These models are strong picks generally but exceed the 18 GB this card makes available.
| Model | Needs at Q4_K_M | Short by |
|---|---|---|
| Qwen 3.7 35B-A3B | 22 GB | ~4 GB |
| Gemma 4 31B | 19.6 GB | ~1.6 GB |
| Nemotron 70B Instruct | 43.5 GB | ~25.5 GB |
| Qwen 3.6 35B-A3B | 22 GB | ~4 GB |
Gemma 4 27B ⭐ is the strongest all-round pick that fits its 18 GB.
Qwen 3.6 27B — 28B parameters, needing 17.6 GB at Q4_K_M.
About 18 GB of its 24 GB, because macOS reserves a share of unified memory for the system.