▸ DEVICE UNDER TEST
ARM Mali-G68 MC2 — 3 GB VRAM.
▸ ARM MALI-G68 MC2 SPEC
- BRAND
- Apple
- VRAM
- 3 GB
- BANDWIDTH
- 14 GB/s
- FP16 COMPUTE
- 0.1 TFLOPS
- FP32 COMPUTE
- 0.05 TFLOPS
- TDP
- 2 W
- ARCHITECTURE
- Valhall
▸ AI CAPABILITY
61/ 331 models @ Q4
With 3 GB VRAM and 14 GB/s bandwidth, this GPU handles models up to 3.2B parameters.
Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~2 tok/s.
§ 01TOP MODELS FOR ARM MALI-G68 MC2
61 FIT · SHOWING 20| MODEL | SIZE | VRAM Q4 | TOK/S | AVG |
|---|---|---|---|---|
| granite-4.0-h-micro 3.2B | 3.2B | 2.4 GB | 4 | 18.4 |
| Llama-3.2-3B | 3.2B | 2.4 GB | 4 | 17.9 |
| Falcon3-3B | 3.1B | 2.4 GB | 4 | 25.7 |
| Qwen 2.5 3B | 3.1B | 2.4 GB | 4 | 37.2 |
| SmolLM3-3B | 3.1B | 2.4 GB | 4 | 30.5 |
| Cogito 3B | 3B | 2.3 GB | 4 | 22.1 |
| Falcon-H1 3B | 3B | 2.3 GB | 4 | 49.5 |
| Ministral 3B | 3B | 2.3 GB | 4 | 29.6 |
| StarCoder2 3B | 3B | 2.3 GB | 4 | 9.5 |
| Dolly v2 3B | 2.8B | 2.2 GB | 4 | 5.6 |
| StableLM Zephyr 3B | 2.79B | 2.2 GB | 4 | 14.9 |
| Zephyr 3B | 2.79B | 2.2 GB | 4 | 14.4 |
| OPT 2.7B | 2.7B | 2.1 GB | 5 | 28.0 |
| Phi-2 2.7B | 2.7B | 2.1 GB | 5 | 24.1 |
| gemma-2-2b | 2.6B | 2.1 GB | 5 | 22.9 |
| CodeGemma 2B | 2.51B | 2.0 GB | 5 | 22.9 |
| EXAONE Deep 2.4B | 2.4B | 2.0 GB | 5 | 27.1 |
| Qwen3.5-2B | 2.3B | 1.9 GB | 5 | 19.3 |
| Qwen2-VL 2B | 2.21B | 1.8 GB | 6 | 28.3 |
| Gemma 1 2B | 2B | 1.7 GB | 6 | 20.2 |