Will DeepSeek’s new model break the ‘long-context’ bottleneck holding back LLMs?
- Posted on October 22, 2025
- By South China Morning Post
- 4 Views
Will DeepSeek’s new model break the ‘long-context’ bottleneck holding back LLMs?
The solution proposed by DeepSeek in its latest paper is to convert text tokens into images, or pixels, using a vision encoder.
