DeepSeek V4 Flash Dominates ARC-AGI Cost-Performance Frontier at 1/4 the Cost

GregKamradt · x · 2026-08-08

DeepSeek V4 Flash has delivered an outstanding performance on the ARC-AGI benchmark, setting a new standard on the cost-to-performance Pareto frontier.

According to verified results from ARC Prize, the model scored 61.4% on ARC-AGI-2 ($0.04/task) and 89.0% on ARC-AGI-1 ($0.02/task). This effectively matches the performance of GPT-5.6 Luna (Max) at just a quarter of the cost.

Related event: DeepSeek V4 Flash Tops ARC-AGI Cost-Performance Frontier(2 posts)→

Original post →

More from Models

Models channel →