Loading…
Loading…
Models / Qwen3.8-Flash-Next
1 model (~180B) · ↓ 3.6M 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 |
|---|---|---|---|---|
| OriginalQwen/Qwen3.8-Flash-Next | vLLM | 360 GB | 403.1 GB | $11.77/h 8x RTX PRO 6000 |
| FP8Qwen/Qwen3.8-Flash-Next-FP8 | vLLM | 186 GB | 209.3 GB | $4.73/h 8x RTX 5090 |
| NVFP4 (RadixArk)RadixArk/Qwen3.8-Flash-Next-NVFP4 | vLLM | 135 GB | 153.3 GB | $1.38/h 8x RTX 3090 |
| NVFP4 (nvidia)nvidia/Qwen3.8-Flash-Next-NVFP4 | vLLM | 133 GB | 150.5 GB | $1.38/h 8x RTX 3090 |
| NVFP4 (local-inference-lab)local-inference-lab/Qwen3.8-Flash-Next-NVFP4 | vLLM | 106 GB | 121.3 GB | $1.38/h 8x RTX 3090 |
| Uncensored FP8 (orcarouter)orcarouter/Qwen3.8-Flash-Next-Uncensored-FP8Licence to accept | vLLM | 186 GB | ~224.9 GB | $4.73/h 8x RTX 5090 |
| JiRackDeltaNet 180b (CMSManhattan)CMSManhattan/JiRackDeltaNet_180b | vLLM | 359 GB | 402.1 GB | $11.77/h 8x RTX PRO 6000 |
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").