DeepSeek V4
LLM with default thinking mode
Episodes
-
#6217: Fine-Tuning a Trillion-Parameter Model at HomeWhat does it actually take to full fine-tune a trillion-parameter model? Sixteen terabytes of optimizer state, hundreds of GPUs, and a very big bill. -
#6211: Attention Is Still Winning: The Hybrid Model CensusMamba layers shipped — inside hybrids. A census of Qwen, Nemotron, Kimi, Granite, and why MiniMax went back to full attention. -
#5673: Why Lowering Temperature Broke Daniel's ScriptsLowering temperature should make output safer, not incoherent. Daniel's DeepSeek scripts say otherwise — and the reason may be stranger than he thi... -
#1606: DeepSeek’s Return: V4, R2, and the AI Pricing WarDeepSeek returns with a trillion-parameter model and rock-bottom pricing. Explore the tech behind V4 and the mystery of the Hunter Alpha leak. -
#258: The Geographic Soul of AI: Mapping the Global Data DivideWhy does an AI see a Chinese supermarket instead of a Western one? Explore how training data shapes the cultural worldview of modern models.