Nathan Lambert Says Strong Teams Can Be Worth More Than Compute for LLMs

What happened

Nathan Lambert argues that a strong organizational structure and team are underrated factors in building a good LLM, and can be worth “multiple multiples in compute.” He says many model gains come from consistent execution and using compute wisely, with some tricks also playing a role.

Read next

• More changes in today’s briefingIntroducing GPT-6.1 Terra

The story so far

Nathan Lambert argued that strong teams and organizational structure are underrated contributors to building a good large language model and can be worth “multiple multiples in compute.” He said many model improvements come from consistent execution and using compute wisely, with some gains also coming from technical tricks.