Might be better than you think
It's good at this task but surprisingly inept at recording
DeepSeek V4.1 Flash might be an absolute speed demon.
⚡ I’m seeing nearly 400 tokens/sec in my video tests. That is wild. I’m running a full benchmark now to see how it holds up across different workloads. I’ll share the numbers as soon as it’s finished. Want to try it
GLM 5.3 Flash vs DeepSeek V4.1, equal conditions, similar time to completion of the scene.
DeepSeek had issues with recording a video (fog + distance), but the actual model is better, so I'd say they are about equal.
You have reached the end of the archive
All of deepseek