Pathway has announced BDH-CQ, a 150 million parameter post-Transformer model that achieves a score of 29.5% on the ARC-AGI-1 benchmark. This model reportedly sets a new cost-efficiency frontier, with a computed cost of only $0.0007 per task. The architecture utilizes recurrent memory and latent reasoning, diverging from traditional long token-based chains of thought. This development is potentially linked to a predicted architectural breakthrough in memory efficiency, with OpenAI researcher Lukasz Kaiser involved as an investor and advisor. AI
IMPACT This model's novel architecture and cost-efficiency could influence future AI development, particularly in memory-intensive tasks.
RANK_REASON The item describes a new model release and its performance on a benchmark, fitting the research category. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
