We gave the exact same Paris café image-generation task to Kimi K3, DeepSeek, GPT-5.6, Gemini 3.8, and GLM.
The results were genuinely surprising. GPT-5.6 and Gemini 3.8 performed poorly. GLM and DeepSeek were decent. But Kimi K3 completely killed the competition. The other
Someone put GPT, Gemini and DeepSeek inside a calculator.
Exams are about to get very interesting.
DeepSeek Harness has officially launched global public preview and is fully open source!
Based on the Cordis "everything is plug-in" architecture, it can be run as a desktop application or Web UI. Supports document processing, intelligent coding and bug fixing, research retrieval and background automation tasks. Just one line of command: npx deepseek-ai/dsh web Experience it now.
You pay for AI APIs by the dollar.
See each balance before it runs out. AI Quota Tool for VS Code now tracks your API keys: DeepSeek, Kimi, OpenRouter, Anthropic and OpenAI. 🧵
You have reached the end of the archive
All of deepseek