Loading…
Loading…
Models / DeepSeek-V4-Flash-0731
1 model (~304B) · ↓ 4.3M downloads a month · from $1.38/h on 8x RTX 3090 right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originaldeepseek-ai/DeepSeek-V4-Flash-0731 | vLLM | 167 GB | 188.5 GB | $2.94/h 2x RTX PRO 6000 |
| keys DeepSeekV4 Flash GA 0731 Dspark Abliterated Anchored Tensors (drowzeys)drowzeys/keys-DeepSeekV4-Flash-GA-0731-Dspark-Abliterated-Anchored-TensorsLicence to accept | vLLM | 167 GB | ~202.5 GB | $4.73/h 8x RTX 5090 |
| spark (0xSero)0xSero/deepseek-v4-flash-0731-spark | vLLM | 107 GB | 121.7 GB | $1.38/h 8x RTX 3090 |
| Abliterated (lovesenko)lovesenko/DeepSeek-V4-Flash-0731-Abliterated | vLLM | 167 GB | 188.5 GB | $2.94/h 2x RTX PRO 6000 |
| NVFP4 (nvidia)nvidia/DeepSeek-V4-Flash-0731-NVFP4 | vLLM | 176 GB | 198.1 GB | $4.73/h 8x RTX 5090 |
Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").