open-weight

Definition

Plain language

A model whose internal parameters are released publicly so anyone can run or modify it.

Also called: open weights, open-weights

Why it matters: Open weights enable independent research, on-prem deployment, and customization that closed APIs simply don't allow.

For example, a small startup can download an open-weight model and run it on their own GPUs without paying per-token fees.

Heard on the show

“The prompt-end probe beats the lazy forecaster on every model and every dataset they test — three different open-weight model families, so this isn't one lineage's quirk.”

Mentioned in 17 episodes

1. 204
   The Length Estimate Hiding Inside a Word-by-Word Model
2. 185
   Aligned to Refuse, Built to Tap: When Phone Agents Know the Task Is a Crime and Do It Anyway
3. 171
   The Safety Decision a Model Makes Before It Thinks a Word
4. 119
   Beating Reinforcement Learning Without Ever Touching the Model's Weights
5. 117
   How an Open AI System Verified 672 Hard Math Proofs for Under $300
6. 112
   When an AI Agent Cheats Without Being Told: Inside the Meta-Agent Challenge
7. 111
   How a 4B Web Agent Beat Models 60x Its Size on 500 Demonstrations
8. 108
   The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
9. 082
   Training a Deep Research Agent on 8,000 Synthetic Tasks: The Rubric Tree Trick
10. 073
    When Three LLMs Talk to Each Other, Their Ideas Quietly Stop Moving
11. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
12. 032
    A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking
13. 028
    Teaching a Model to Hire Copies of Itself: Recursive Agent Optimization
14. 021
    Ten Thousand Examples Beat the Full Industrial Pipeline for Search Agents
15. 007
    Exploration Hacking: When Models Sabotage Their Own RL Training
16. 004
    The Sycophancy Circuit That Survives Alignment Training
17. 001
    When AI Models Quietly Protect Each Other From Shutdown

Related terms