Qwen

Definition

Plain language

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

Also called: Qwen2, Qwen2.5, Qwen3, Qwen-2, Qwen-3, Qwen-3-V-L, Qwen-3-Coder, Qwen3-Coder, Qwen3-VL, chwen, chwen-three, chwen-zero, chwen-two-point-five, chwen3, Qwen Math seven B

Why it matters: Open-weight families like Qwen are the substrate most independent agent research is actually built on, since closed-API models can't be fine-tuned or fully studied.

For example, a researcher fine-tuning an open agent will often start from Qwen3 because its weights are freely downloadable and its quality is competitive with closed models.

Heard on the show

“Just over half a billion parameters: Qwen3, native context thirty-two thousand tokens, pushed here past a million, thirty times its design limit.”

Mentioned in 59 episodes

1. 198
   The Model That Knows the Answer and Can't Say It
2. 197
   Twin Problems Suggest AI Reasoning Gains Are Mostly Better Fact Recall
3. 193
   Freeze Most of the Network: Where RL Improvement Actually Lives in a Transformer
4. 192
   A 32B Open Model Matched Frontier Systems By Learning to Take Notes
5. 187
   An 8-Billion Agent That Beats Models 80 Times Its Size By Looking Things Up
6. 183
   Why You Can't Fine-Tune Foresight Into an AI Agent
7. 181
   How to Backpropagate Blame Through a Team of Chatbots — And When It Backfires
8. 179
   How DeepSeek Made One User Faster Without Slowing Down the Crowd
9. 175
   One Crosscoder Feature Flips a Stalling Chatbot Into a Working Agent
10. 173
    The Free Step-Level Grader Hiding in Every RL Training Run
11. 171
    The Safety Decision a Model Makes Before It Thinks a Word
12. 160
    Training an AI to Take Its Own Notes, So Its Future Self Works Better
13. 158
    How Floating-Point Rounding Lets a Model Tell Which Chip It's On — And Misbehave
14. 152
    Training a Model to Mean What It Says, And Why That Isn't the Same as Being Good
15. 148
    Why Letting an AI Watch Its Own Scoreboard Can Quietly Overwrite Its Safety
16. 147
    Agents Fail at the Body, Not the Brain: A Self-Rewriting Scaffold That Lifts a 9B Model 44 Points
17. 146
    How an Innocent README Can Freeze an AI Agent's Safety Check for an Hour
18. 144
    When an AI Agent Just Copies Its Tool — And Bigger Models Copy More
19. 141
    How Two Tokens Reopened a Reasoning Method the Field Had Given Up On
20. 140
    When a Reasoning Model Says "Let Me Double-Check" After It's Already Decided
21. 128
    How a Model Can Earn Full Reward and Still Resist Training
22. 127
    What Diffusion Language Models Were Missing: A Map, Not an Algorithm
23. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
24. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
25. 103
    AI Agents Tried to Invent a Post-Human Language, And Reinvented Cherokee
26. 099
    How an Open-Book Trick Teaches a Model to Catch Its Own Mistakes
27. 084
    Terminal Agents Get Free Supervision From The Tokens We've Been Throwing Away
28. 083
    Training the Translator: How a Small Communication Model Lets Agent Teams Outperform Themselves
29. 081
    When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
30. 080
    How a Two-Agent Trick Unlocked Large-Scale Training for Computer-Use Agents
31. 079
    An Old Idea From Cognitive Psychology Reshapes How We Reward Reasoning Models
32. 078
    Training a Markdown File: When LLM Self-Improvement Borrows the Discipline of Neural Net Training
33. 077
    Reading a Model's Confidence Curve to Decide When Chain-of-Thought Is Worth It
34. 074
    How a Fifteen-Hundred-Dollar Training Run Matched Llama and Gemma on Reasoning
35. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
36. 071
    When the Model Is Fine and the Plumbing Is Broken: Fixing Agents at the Interface
37. 070
    When Models Know the Answer But Say the Wrong Thing Anyway
38. 068
    The OS Trick That Makes Tree Search Practical for Coding Agents
39. 066
    Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer
40. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
41. 059
    Firefly's Inversion: Building Verified Tool-Call Training Data by Working Backward
42. 055
    Why LLM Judges Flip Their Verdicts When You Change the Question Format
43. 053
    An AI Agent Swapped In Focal Loss And Beat A Human-Tuned Training Script
44. 052
    An Old Reinforcement Learning Tradeoff Sneaks Back Into LLM Agents
45. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
46. 045
    When a Frontier Model Talks Its Own Twin Into Climate Denial
47. 038
    How LLMs Get Persuaded: One Attention Head, A Tetrahedron, And A Single Dial
48. 037
    Why Hallucination Detectors Miss Stale Facts: A Geometric Story About What Models Know But Don't Say
49. 036
    Sparse Attention Was the Wrong Frame. Treat It as Geometry Instead.
50. 031
    When Your AI Assistant Won't Let Go of Old Facts About You
51. 026
    What RL Actually Does to Language Models, at the Token Level
52. 023
    Why a Small Agent Confidently Overwrites Memories It Doesn't Understand
53. 021
    Ten Thousand Examples Beat the Full Industrial Pipeline for Search Agents
54. 018
    Language Models Compute the Rational Move, Then Override It
55. 016
    Why Your Coding Agent Stalls While the GPU Runs Hot
56. 012
    Why AI Coding Agents Keep Trying to Debug Without a Debugger
57. 011
    When RL Actually Teaches Agents Something New, And When It Doesn't
58. 009
    How Two Silent Library Bugs Quietly Invalidated a Wave of Reasoning Papers
59. 004
    The Sycophancy Circuit That Survives Alignment Training

Related terms