Greenblatt: 3x jump in model depth points to architecture changes, not scaling
RyanGreenblatt · x · 2026-09-04
Pushing back on a claim that a new model's reasoning depth is "within a factor of 2 of GPT-4," Ryan Greenblatt argues the gain is mostly architecture-driven: open-weight scaling trends suggest a 3x depth increase would normally require 81x more parameters (equivalent to >6000x more compute), so something beyond normal scaling must be at play.
More from Models
- Opus 5.5 tops OpenRouter spending in 20 of 29 task categories — mmurph · 2026-10-21
- Meta team builds open-source 30B LLM from scratch in months, hails 'second era' — TimDarcet · 2026-10-21
- DFlash drafter for Muse Glimmer-30B: 5 layers, blockwise 16-token spec decoding — TimDarcet · 2026-10-21
- EvoPress quantization shrinks 30B model from 56GB to 19GB at near-bf16 accuracy — TimDarcet · 2026-10-21
- Radar chart: Muse Glimmer-30B benchmarked against Gemma4-31B and Qwen3.6-27B — TimDarcet · 2026-10-21
- Privacy evals: Muse Glimmer-30B leaks less than Qwen3.6-27B on CIMemories — TimDarcet · 2026-10-21