150M-parameter reasoning model sets new cost-accuracy frontier on ARC-AGI-1

Thinking Machines' 150M-parameter BDH-CQ reasoning model, which combines in-context learning with recurrent latent reasoning, achieves 29.5% pass@2 on the public ARC-AGI-1 evaluation set at a computed inference cost of $0.0007 per task, breaking the previously reported cost-accuracy Pareto frontier and establishing a new state of the art in benchmark cost efficiency.

Collection Papers that made me appreciate my major and my life a little more. obs=Observation, innov=Innovation. Most papers are abt improving tiny models. • 268 items • Updated • 65 Abstract A 150M-parameter reasoning model using recurrent latent reasoning and in-context learning achieves a new cost-accuracy frontier on ARC-AGI-1. thinkingmachines/Inkling-Small /thinkingmachines/Inkling-Small We introduce BDH-CQ, a reasoning model that combines in-context learning /papers?q=in-context%20learning with recurrent latent reasoning /papers?q=recurrent%20latent%20reasoning . Inputs presented at inference time continuously update the model's recurrent memory; the model then solves a query through iterative computation in a high-dimensional latent space /papers?q=high-dimensional%20latent%20space , without verbalizing its intermediate reasoning. We evaluate the model on the public ARC-AGI-1 /papers?q=ARC-AGI-1 evaluation set and use controlled ARC-like interventions to study what it learns from demonstrations, how consistently it applies an inferred transformation, and which concepts remain difficult. A 150M-parameter configuration reaches 29.5% pass@2 at a computed inference cost of \$0.0007 per task. This operating point breaks through the previously reported ARC-AGI-1 /papers?q=ARC-AGI-1 cost-accuracy Pareto frontier /papers?q=cost-accuracy%20Pareto%20frontier , establishing a new state of the art in benchmark cost efficiency. Community This is an automated message from the Librarian Bot https://huggingface.co/librarian-bots . I found the following papers similar to this paper. The following papers were recommended by the Semantic Scholar API Penelope: Localized Latent Recurrence for Efficient Structured Reasoning https://huggingface.co/papers/2607.25915 2026 Recursive Vision Language Models for General Symbolic Reasoning https://huggingface.co/papers/2608.01534 2026 DiscoLoop: Looping Discrete Embeddings and Continuous Hidden States for Multi-hop Reasoning https://huggingface.co/papers/2607.00341 2026 J-CoT: Chain-of-Thought in J-Space https://huggingface.co/papers/2607.21981 2026 Learning to Refine Hidden States for Reliable LLM Reasoning https://huggingface.co/papers/2606.17524 2026 Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers https://huggingface.co/papers/2606.31779 2026 What Survives When You Compress a Recursive Reasoner for the Edge? https://huggingface.co/papers/2606.26488 2026 Please give a thumbs up to this comment if you found it helpful If you want recommendations for any Paper on Hugging Face checkout this https://huggingface.co/spaces/librarian-bots/recommend similar papers Space You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend Get this paper in your agent: hf papers read 2608.09888 Don't have the latest CLI? curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0 No model linking this paper Datasets citing this paper 0 No dataset linking this paper Spaces citing this paper 0 No Space linking this paper Collections including this paper 5 Browse 5 collections that include this paper /collections?paper=2608.09888