GPU ARuns more models
Apple M2 Ultra (192GB)
- VRAM
- 192 GB
- Bandwidth
- 800 GB/s
- Street price
- $5,499
- Vendor
- apple
GPU B
NVIDIA RTX 4090
- VRAM
- 24 GB
- Bandwidth
- 1008 GB/s
- Street price
- $1,999
- Vendor
- nvidia
The short answer
Apple M2 Ultra (192GB) can run 13 models that NVIDIA RTX 4090 can't fit in VRAM — mostly the larger models. For the 33 models both can handle, speeds are similar. If you want headroom for bigger models, Apple M2 Ultra (192GB) is the clear choice.
13 models A runs, B can't
33 models B is 20%+ faster
0 equal (both run, <20% diff)
8 too large for either
Model-by-model fit
Click any row for the full breakdown. Tie% shown in the Winner column when both GPUs run the model within 20% of each other.
| Model | Apple M2 Ultra (192GB) | NVIDIA RTX 4090 | Winner |
|---|---|---|---|
c4ai-command-r-v01 35B 35B · command | 10 tok/s · FP16 | 42 tok/s · Q4_K_M | B (faster) |
Command-R+ 104B 104B · command | 6 tok/s · Q8_0 | Too large | A (only) |
DeepSeek R1 Distill Llama 8B 8B · deepseek | 41 tok/s · FP16 | 52 tok/s · FP16 | B (faster) |
DeepSeek R1 Distill Qwen 14B 14.8B · deepseek | 23 tok/s · FP16 | 56 tok/s · Q8_0 | B (faster) |
DeepSeek R1 Distill Llama 70B 70.6B · deepseek | 5 tok/s · FP16 | Too large | A (only) |
DeepSeek R1 671B 671B · deepseek | Too large | Too large | — |
DeepSeek-V3 685B 685B · deepseek | Too large | Too large | — |
DeepSeek-V3.2 685.4B 685.4B · deepseek | Too large | Too large | — |
gemma-2-9b 9.2B · gemma | 36 tok/s · FP16 | 45 tok/s · FP16 | B (faster) |
gemma-2-27b 27.2B · gemma | 12 tok/s · FP16 | 38 tok/s · Q6_K | B (faster) |
Llama 3.1 8B Compact 8B · llama | 37 tok/s · FP16 | 47 tok/s · FP16 | B (faster) |
CodeLlama 34B 34B · llama | 10 tok/s · FP16 | 44 tok/s · Q4_K_M | B (faster) |
CodeLlama 34B 34B · llama | 10 tok/s · FP16 | 44 tok/s · Q4_K_M | B (faster) |
Llama 3.3 70B 70.6B · llama | 5 tok/s · FP16 | Too large | A (only) |
Llama 3.1 70B 70.6B · llama | 5 tok/s · FP16 | Too large | A (only) |
Llama 4 Scout 17B 109B · llama | 5 tok/s · Q8_0 | Too large | A (only) |
Llama-4-Maverick-17B-128E 400B · llama | Too large | Too large | — |
Llama 3.1 405B 405B · llama | Too large | Too large | — |
Mistral 7B v0.1 7.25B · mistral | 45 tok/s · FP16 | 57 tok/s · FP16 | B (faster) |
Codestral 22B 22.2B · mistral | 15 tok/s · FP16 | 46 tok/s · Q6_K | B (faster) |
Mixtral 8x7B Instruct v0.1 47B · mixtral | 6 tok/s · FP16 | Too large | A (only) |
Mistral Large 2 123B 123B · mistral | 5 tok/s · Q8_0 | Too large | A (only) |
Phi-4-mini 3.8B 3.8B · phi | 84 tok/s · FP16 | 106 tok/s · FP16 | B (faster) |
Phi-4 14B 14B · phi | 21 tok/s · FP16 | 53 tok/s · Q8_0 | B (faster) |
Qwen 2.5 1.5B 1.5B · qwen | 195 tok/s · FP16 | 246 tok/s · FP16 | B (faster) |
Qwen 2.5 3B 3.1B · qwen | 102 tok/s · FP16 | 128 tok/s · FP16 | B (faster) |
Qwen3.5-4B 4.7B · qwen | 69 tok/s · FP16 | 87 tok/s · FP16 | B (faster) |
Qwen 2.5 7B 7.6B · qwen | 43 tok/s · FP16 | 55 tok/s · FP16 | B (faster) |
Qwen 2.5 7B 7.6B · qwen | 43 tok/s · FP16 | 55 tok/s · FP16 | B (faster) |
Qwen 3 8B 8B · qwen | 37 tok/s · FP16 | 47 tok/s · FP16 | B (faster) |
Qwen3.5-9B 9.7B · qwen | 34 tok/s · FP16 | 43 tok/s · FP16 | B (faster) |
Qwen 3 32B 32B · qwen | 9 tok/s · FP16 | 41 tok/s · Q4_K_M | B (faster) |
Qwen3.5-35B-A3B 36B · qwen | 9 tok/s · FP16 | 41 tok/s · Q4_K_M | B (faster) |
Qwen 2.5 72B 72.7B · qwen | 5 tok/s · FP16 | Too large | A (only) |
Qwen 2.5 72B 72.7B · qwen | 5 tok/s · FP16 | Too large | A (only) |
Llama 3.2 1B 1.24B · llama | 238 tok/s · FP16 | 300 tok/s · FP16 | B (faster) |
Llama 4 Scout 17B 109B · llama | 5 tok/s · Q8_0 | Too large | A (only) |
DeepSeek R1 671B 671B · deepseek | Too large | Too large | — |
Gemma 3 27B 27B · gemma | 11 tok/s · FP16 | 40 tok/s · Q5_K_M | B (faster) |
Qwen 3 8B 8B · qwen | 37 tok/s · FP16 | 47 tok/s · FP16 | B (faster) |
Qwen 3 32B 32B · qwen | 9 tok/s · FP16 | 41 tok/s · Q4_K_M | B (faster) |
Llama 3.1 8B Compact 8B · llama | 37 tok/s · FP16 | 47 tok/s · FP16 | B (faster) |
Mixtral 8x7B Instruct v0.1 47B · mixtral | 6 tok/s · FP16 | Too large | A (only) |
Mistral Small 3.2 24B 24B · mistral | 15 tok/s · FP16 | 45 tok/s · Q6_K | B (faster) |
Command A 111B 111B · command | 6 tok/s · Q8_0 | Too large | A (only) |
DeepSeek R1 0528 685B · deepseek | Too large | Too large | — |
DeepSeek-V3-0324 684.5B · deepseek | Too large | Too large | — |
DeepSeek-R1-0528-Qwen3-8B 8.2B · qwen | 42 tok/s · FP16 | 53 tok/s · FP16 | B (faster) |
Qwen3-235B-A22B-Instruct-2507 235B · qwen | 5 tok/s · Q4_K_M | Too large | A (only) |
Qwen3-30B-A3B-Instruct-2507 30B · qwen | 23 tok/s · Q8_0 | 52 tok/s · Q4_K_M | B (faster) |
Qwen3-4B-Instruct-2507 4B · qwen | 87 tok/s · FP16 | 110 tok/s · FP16 | B (faster) |
gemma-4-E4B-it 8B · gemma | 44 tok/s · FP16 | 55 tok/s · FP16 | B (faster) |
gemma-4-26B-A4B-it 26.5B · gemma | 26 tok/s · Q8_0 | 41 tok/s · Q6_K | B (faster) |
gemma-4-31B-it 32.7B · gemma | 21 tok/s · Q8_0 | 48 tok/s · Q4_K_M | B (faster) |
Want a different pairing? Browse all comparisons →