Garry Tan calls for US open-weight models distilled from frontier systems

single source · 1 articles · policy · confidence: low · first seen 2026-09-11 20:59 UTC

On 11 September 2026, Garry Tan of Y Combinator called for smaller American open-weight AI labs to distil US frontier models — that is, train a smaller model to reproduce a larger model's outputs — rather than depend on Chinese models. Tan argues this would give the US a more robust set of open-weight options. The TechCrunch report did not name specific labs, models, or policy changes.

What this means for you

Nothing to do today. Tan has expressed a preference, not announced a programme or policy change. No specific labs or models are named in the report.

Key facts

  • ·Garry Tan is affiliated with Y Combinator. source
  • ·Tan called for smaller American open-weight AI labs to use distillation training techniques on US frontier AI models. source
  • ·The stated goal is to give the US robust open-weight options that are not Chinese. source

What the sources say

  • TechCrunchReports Tan's call for US labs to distil frontier models into non-Chinese open weights.

Sources

The original reporting. Follow these — they did the work.

← the wire