On September 20, xAI announced "Grok 4 Fast," a reasoning model specialized for cost efficiency. The company explained that it maximized intelligence density through large-scale reinforcement learning and stated that it achieved performance equivalent to Grok 4 on benchmarks using an average of 40% fewer thought tokens. Combined with a decrease in token unit prices, the company claims that the cost to achieve the same performance is reduced by 98%.
According to verification by the independent review organization Artificial Analysis, the model demonstrated the SOTA price-to-intelligence ratio among public models. Furthermore, it is reported to feature a 2-million-token context window, an architecture integrating reasoning and non-reasoning modes, and agentic search capabilities driven by tool-use reinforcement learning.
Sources: Grok 4 Fast (HN 96pt, 76 comments) (HN Search (backfill))