We gave China's top open source models 15 minutes to draw a pelican riding a bicycle.
DeepSeek needed two tries to get there. MiniMax barely scraped across the finish line. GLM spent the entire 15 minutes "thinking" and never even started drawing. So what did GPT do? Check here
DeepSeek Harness vs Claude Code.
We gave both the EXACT same prompt: "Build a 3D animated website and add a working Tetris game." One finished in 11 minutes. The other was still working after 30. But speed wasn't the whole story.
Deepseek v4 flash vision graphic and text creation evaluation
For the first time, I connected the deepseek v4 flash vision model to my website. The modal proxy performance is close to Opus-4.8, making it easier to create graphics and text.
Mr Dereck Goto sits down with CGTN's Gao Junya to discuss AI's double-edged sword for journalism and…
Mr Dereck Goto sits down with CGTN's Gao Junya to discuss AI's double-edged sword for journalism and why Chinese models like DeepSeek could level the playing field for African media
You have reached the end of the archive
All of deepseek