Llama

Definition

Plain language

Meta's family of open-weight large language models.

Also called: Llama-2, Llama-3, Llama-3.1, Llama-3.3, Llama-4

Why it matters: The Llama family is a backbone of open AI research, so its capabilities and licensing terms shape what independent labs and startups can build.

For example, a researcher might fine-tune Llama-3-8B on medical Q&A to build a specialized assistant without paying for proprietary model access.

Heard on the show

“Across GPT-4, Llama, Mistral, and Gemma, the hedged prompts reliably got responses that were more readable — that one moves in the women-associated direction in nearly every model-and-document cell.”

Mentioned in 36 episodes

1. 247
   One Edited Photo, an Honest Caption, and a RAG System That Believes It
2. 241
   Swapping the Name Did Nothing, But Hedging Moved Every Model
3. 231
   Silencing a Chatbot's 'I'm Conscious' Quietly Rewires Its Whole Worldview
4. 230
   Why AI Survey Panels Break Before the Dice Ever Roll
5. 225
   How a Frozen Model Went From Zero to Sixty Percent by Borrowing Another's Thinking
6. 211
   The AI Watchdog That Approved More Cheating When It Could Read Minds
7. 207
   An AI Graded Its Own Math Test 94 Percent — It Actually Scored 20
8. 183
   Why You Can't Fine-Tune Foresight Into an AI Agent
9. 182
   How a Tiny Model Too Weak to Plan Cuts a Bigger Agent's Hallucinations by 80%
10. 181
    How to Backpropagate Blame Through a Team of Chatbots — And When It Backfires
11. 152
    Training a Model to Mean What It Says, And Why That Isn't the Same as Being Good
12. 148
    Why Letting an AI Watch Its Own Scoreboard Can Quietly Overwrite Its Safety
13. 119
    Beating Reinforcement Learning Without Ever Touching the Model's Weights
14. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
15. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
16. 107
    How a Market of Crippled AI Agents Outscored One Unrestricted Model
17. 087
    When No Agent Reads the Whole Document: A Universal Cliff in Multi-Agent Review
18. 077
    Reading a Model's Confidence Curve to Decide When Chain-of-Thought Is Worth It
19. 074
    How a Fifteen-Hundred-Dollar Training Run Matched Llama and Gemma on Reasoning
20. 073
    When Three LLMs Talk to Each Other, Their Ideas Quietly Stop Moving
21. 071
    When the Model Is Fine and the Plumbing Is Broken: Fixing Agents at the Interface
22. 070
    When Models Know the Answer But Say the Wrong Thing Anyway
23. 069
    When Smarter Models Forecast Worse: The Hidden Failure Mode in LLM Predictions
24. 058
    Why Upgrading Your AI Auditor to a Smarter Model Can Make Your System Less Safe
25. 055
    Why LLM Judges Flip Their Verdicts When You Change the Question Format
26. 053
    An AI Agent Swapped In Focal Loss And Beat A Human-Tuned Training Script
27. 038
    How LLMs Get Persuaded: One Attention Head, A Tetrahedron, And A Single Dial
28. 037
    Why Hallucination Detectors Miss Stale Facts: A Geometric Story About What Models Know But Don't Say
29. 027
    When AI Agents Build the Serving Stack: A Bet on Bespoke Infrastructure
30. 025
    The Missing Gradient Term That Predicts Sycophancy in RLHF
31. 019
    When the Best Reward Model Trains the Worst Policy: Inside EvoLM
32. 018
    Language Models Compute the Rational Move, Then Override It
33. 015
    The Audit Number Isn't What You Think: Sycophancy and the Case Against Single-Prompt Bias Tests
34. 009
    How Two Silent Library Bugs Quietly Invalidated a Wave of Reasoning Papers
35. 004
    The Sycophancy Circuit That Survives Alignment Training
36. 001
    When AI Models Quietly Protect Each Other From Shutdown

Related terms