18 August 2026

Smaller AI models match larger ones using hidden reasoning and memory

• A smaller model called BDH-CQ achieved 29.5% accuracy on ARC-AGI, a benchmark for general reasoning, using internal reasoning steps and temporary memory storage.
• GPT-5.6 Sol improved from 13.3% to 38.3% on the same benchmark by keeping reasoning steps and using 6 times fewer input tokens than before.
• Both examples show that how a model thinks internally, not just its size, determines how well it solves problems.

How it was covered

Latent Spaceswyx & Alessio

Small models like BDH-CQ use latent-space reasoning with temporary memory to hit 29.5% on ARC-AGI at very low cost, while GPT-5.6 Sol improved from 13.3% to 38.3% on ARC-AGI-3 using retained reasoning and 6x fewer tokens. The newsletter frames memory and compaction strategy as a first-class capability multiplier.