DeepSeek V4 Flash Vision Uncensored
DeepSeek V4 Flash Multimodal T5000 Optimized Version
Description
1. Single stream decoding 37.40 ~ 39.29 TPS, 8 concurrent decoding 122.01 ~ 124.53 TPS 2. Prefill 674.70 TPS（18K）/ 268.87 TPS（1M） 3. TTFT 5K 7.22 ~ 7.25s 4. Up to 1M context 5. Two T5000 computing cabins are needed 6. Unreviewed model
Screenshots
Mobile Screenshots
App Information
Original App