Gopher 280B: DeepMind officially enters the LLM race

In one sentence DeepMind releases Gopher, a 280B dense model, alongside a systematic 152-task study and a companion paper on ethical considerations of foundation models.

Verified Official source

ShareLinkedIn X

DeepMind, until then focused on AlphaGo, AlphaFold and reinforcement learning, enters the large language model arena with Gopher, 280 billion parameters.

It's not just a model announcement: there's also a scientific study comparing Gopher against GPT-3 and others on 152 tasks. It shows where scale helps (reading, knowledge) and where it doesn't (logic, math) — an observation central to the years that follow.

DeepMind also publishes a companion paper on ethical risks, signaling a more cautious posture than OpenAI.