/Tech1h ago

Nathan Lambert, AI2 post-training lead, and Finbarr Timbers argue frontier training is shifting from DeepSeek-R1 toward expert-based distillation

Story Overview

Nathan Lambert and Finbarr Timbers spend nearly an hour dissecting how top labs now blend outputs from several specialized models instead of leaning on the single-teacher reinforcement patterns that defined DeepSeek-R1. Their exchange highlights concrete tweaks that could help open pipelines like OLMo close the gap, while leaving exact compute budgets and training curves as open variables.

9125128913.2K

#80

Original post

Nathan Lambert@natolambert#80inTech

New podcast with @finbarrtimbers! We survey the latest post-training recipes, from GLM 5.1, Kimi K2.6, DeepSeek V4, Xiaomi MiMo V2.5, Nemotron Ultra, etc. and discuss: - Why the industry slowly shifted to multi-teacher on-policy distillation (MOPD). - What an Olmo-style recipe would need improvements in - How post-training works / suits larger organizational efforts - Career advice in the foothills of the singularity - and other topics

I heard y'all wanted me to start doing this, so making some time when I'm in funemployment!

Chapters:

00:00 Introduction & Olmo reflections 06:28 Post-train recipes review (history) 23:00 2026’s model recipes (MiMo Flash, DeepSeek V4, GLM 5, Kimi K2.6, etc.) 39:05 Open-ended post-training discussions 48:22 Career advice in the LLM race

Links below, please follow @interconnectsai and like and subscribe and buy my book?

6:44 AM · Jun 16, 2026 · 10.7K Views

Industry Shift

Multi-teacher methods may soon become table stakes

The pair maps the move toward multi-teacher on-policy distillation across GLM-5, Kimi K2.6, and MiMo Flash, noting that single-model supervision appears to be losing favor at the frontier.

Open Question

Open OLMo recipes still need clearer scaling math

Lambert and Timbers flag specific places where current OLMo post-training steps could borrow from the new distillation patterns, yet they stop short of releasing any updated training schedules or token counts.

Sentiment

Users express strong excitement for the podcast surveying 2026 AI post-training recipes and multi-teacher distillation, praising the hosts and indicating they must listen soon.

Pos

100.0%

Neg

0.0%

4 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS2KREPLIES1

Nathan Lambert@natolambert

@finbarrtimbers YouTube: https://www.youtube.com/watch?v=sbXEPxIazqY&list=PLL1tdVxB1CpVpEtMHxwuR4uI4Lxjw00_y&index=10

Nathan Lambert@natolambert

I heard y'all wanted me to start doing this, so making some time when I'm in funemployment!

Chapters:

Links below, please follow @interconnectsai and like and subscribe and buy my book?

1h2K95

BOOKMARKS6LIKES18

finbarr@finbarrtimbers

Fun chat I had with Nathan last week!

Nathan Lambert@natolambert

I heard y'all wanted me to start doing this, so making some time when I'm in funemployment!

Chapters:

Links below, please follow @interconnectsai and like and subscribe and buy my book?

1h1.4K186

Nathan Lambert@natolambert

@finbarrtimbers Interconnects: https://www.interconnects.ai/p/frontier-post-training-recipe-review

Nathan Lambert@natolambert

@finbarrtimbers YouTube: https://www.youtube.com/watch?v=sbXEPxIazqY&list=PLL1tdVxB1CpVpEtMHxwuR4uI4Lxjw00_y&index=10

1h1K41

Saurabh Shah@saurabh_shah2

@natolambert @finbarrtimbers Yayyy

21m291

Alex Weers@a_weers

@natolambert @finbarrtimbers 🔥

1h191

Glenn Matlin@GlennMatlin

@natolambert @finbarrtimbers Open ended?? Did you say… open ended??? 👀😍 definitely must listen this week when I hit my AI usage caps and need my fix

17m10

Vivek@vivek_2332

@natolambert @finbarrtimbers awesome!!

1h5