NVIDIA GeForce RTX 4090 vs NVIDIA GeForce RTX 5070 Ti
For running models locally, priced in India. At 32k context the first runs 26 models fully in memory, the second 13.
Specifications
| Spec | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 5070 Ti |
|---|---|---|
| GPU | GeForce RTX 4090 | GeForce RTX 5070 Ti |
| Memory | 24 GB GDDR6X | 16 GB GDDR7 |
| Memory bus | 384-bit | 256-bit |
| Memory bandwidth | 1,008 GB/s | 896 GB/s |
| FP16 compute | 165 TFLOPS | 88 TFLOPS |
| Software | CUDA | CUDA |
| Board power | 450 W | 300 W |
| PCIe generation | Gen 4 | Gen 5 |
| PCIe lanes | x16 | x16 |
What each runs
Estimated decode speed with the most precise quant that fits each one est.
| Model | NVIDIA GeForce RTX 4090 | NVIDIA GeForce RTX 5070 Ti |
|---|---|---|
| Qwen 3.8 27B | 28tok/sUD-Q5_K_M | offload |
| Qwen 3.6 27B | 29tok/sQ5_K_M | offload |
| Gemma 4 31B | 29tok/sQ4_K_M | offload |
| Qwen 3.6 35B-A3B | 145tok/sUD-Q4_K_M | offload |
| Qwen 3.5 27B | 29tok/sQ5_K_M | offload |
| Gemma 4 26B-A4B | 84tok/sUD-Q6_K | offload |
| Qwen 3.5 35B-A3B | 146tok/sQ4_K_M | offload |
| GLM 4.7 Flash | 70tok/sQ5_K_M | offload |
| Devstral Small 2 24B | 28tok/sQ5_K_M | offload |
| gpt-oss 20B | 116tok/sMXFP4 | 103tok/sMXFP4 |
| Qwen3 Coder 30B-A3B | 74tok/sQ4_K_M | offload |
| Gemma 4 12B | 46tok/sQ8_0 | 41tok/sQ8_0 |
| Qwen 3.5 9B | 59tok/sQ8_0 | 52tok/sQ8_0 |
| Mistral Small 3.2 24B | 28tok/sQ5_K_M | offload |
| Gemma 3 27B | 28tok/sQ5_K_M | offload |
| Ministral 3 14B | 32tok/sQ8_0 | 37tok/sQ5_K_M |
| Qwen3 14B | 30tok/sQ8_0 | 39tok/sQ4_K_M |
| DeepSeek R1 0528 Qwen3 8B | 46tok/sQ8_0 | 41tok/sQ8_0 |
| Qwen3 8B | 46tok/sQ8_0 | 41tok/sQ8_0 |
| Phi-4 14B | 28tok/sQ8_0 | offload |
| Qwen 3.5 4B | 112tok/sQ8_0 | 99tok/sQ8_0 |
| Ministral 3 8B | 46tok/sQ8_0 | 41tok/sQ8_0 |
| Granite 4.2 8B | 43tok/sQ8_0 | 38tok/sQ8_0 |
| Gemma 4 E4B | 68tok/sQ8_0 | 61tok/sQ8_0 |
| LFM2.5 8B-A1B | 140tok/sQ8_0 | 125tok/sQ8_0 |
| Llama 3.1 8B | 57tok/sQ6_K | 51tok/sQ6_K |
| Qwen 3.5 122B-A10B | offload | offload |
| Qwen3 Coder Next | offload | won't run |
| Mistral Small 4 119B | offload | offload |
| GLM 4.5 Air | offload | won't run |
| Qwen3 Next 80B-A3B | offload | won't run |
| Nemotron 3.5 Lightning 30B-A3B | offload | offload |
| Kimi Linear 48B-A3B | offload | offload |
| Qwen3 32B | offload | offload |
| DeepSeek R1 Distill Qwen 32B | offload | offload |
| DeepSeek V4 Flash | won't run | won't run |
| GLM 5.3 Flash | won't run | won't run |
| Qwen 3.8 Flash Next | won't run | won't run |
| MiniMax M2.7 | won't run | won't run |
| Nemotron 3 Super 120B-A12B | won't run | won't run |
| gpt-oss 120B | won't run | won't run |
| Llama 3.3 70B | won't run | won't run |