Tag
The Qwen4 family of AI models, including Qwen4-Max, Qwen4-Flash&Qwen4-Plus, and Qwen4-27B, is coming soon with unprecedented parameter scales at the 5T to 10T level, previewed at the Yunqi Conference.
DeepSeek is training a 2 trillion-parameter AI model and plans to eventually build an 8 trillion-parameter model.
Santiago Valdarrama reports that Kimi K3 is the best open-weight model he has tested, a 2.8T-parameter vision model with tool calling, reasoning, and a 1M context window.
Explores whether a 35B parameter model could surpass a 1000B model by scaling its search horizon at test time, using structured process feedback rather than brute-force parameter scaling.