Stop Obsessing Over Small Model Releases
ParaboloidalCrest · reddit · 2026-07-19
The author argues that **pushing for "small models" is somewhat pointless**: if labs have plans to release them, they will eventually; if they don't, they won't change their minds based on public demand. They also point out that teams like Qwen, GLM, Kimi, and DeepSeek no longer need to rely on frequently launching experimental small models to capture market share. Looking ahead, what's more noteworthy might be models from newer labs that are close to being runnable on **consumer-grade hardware**. The post concludes by reminding readers not to ignore lesser-known models, specifically naming **JoyAI's 49B** and the still-in-use **SeedOSS 36B**, praising the latter's performance as "awesome".
More from AGI Musings
- What if the brain could add new words every day and keep expanding thought? — ZeroStateReflex · 2026-07-21
- MIRI’s Plan S is said to have a 66% win probability, far above Plan A’s 18% — eli_lifland · 2026-07-21
- Open-source robotics should not be slowed to protect two labs’ evals — RileyRalmuto · 2026-07-21
- By 2030, an AI-made song by a dead musician could top the charts — VraserX · 2026-07-21
- Garry Tan says AI is making everyone equal as superintelligence reshapes product building — garrytan · 2026-07-21
- Veteran programmer says five waves of automation only made coders more valuable — ZeeshanZiaML · 2026-07-21