xAI has upgraded their fast, cheap, agentic focused model to Grok 4.1 Fast, jumping +4 points on the Artificial Analysis Intelligence Index Key takeaways: ➤ Leading agentic performance: Grok 4.1 Fast with reasoning takes the top spot on our 𝜏²-Bench Telecom Leaderboard with a score of 93.3%, just ahead of Kimi K2 Thinking, and represents a +27 point increase over Grok 4 Fast ➤ Intelligence Pareto frontier: Like its predecessor, Grok 4.1 Fast offers leading intelligence at low cost - it sits only 1 point behind Grok 4 on the Intelligence Index with a ~30x lower per-token cost. This places Grok 4.1 Fast firmly in the most attractive quadrant of our Intelligence Index versus cost comparison ➤ Clear incremental upgrade: Alongside these improvements for agents and tool calling, Grok 4.1 Fast retains the strengths of its predecessor: low pricing ($0.2/0.5 per million tokens input/output), large context window (2 million tokens), and token efficiency Congratulations to the @xai team and @elonmusk on this new release! See below for further analysis

Grok 4.1 Fast achieves a 64 on the Artificial Analysis Intelligence Index, 4 points higher than Grok 4, making it a clear upgrade over its predecessor

Grok 4.1 Fast took 71M tokens to complete the Artificial Analysis Intelligence Index evaluations at a cost of $45, similar to Grok 4 and significantly less on both accounts than most other leading models

Individual benchmark results: MMLU-Pro, GPQA Diamond, Humanity's Last Exam, LiveCodeBench, SciCode, AIME 2025, IFBench, AA-LCR, Terminal-Bench Hard, 𝜏²-Bench Telecom

See our models page for further analysis of performance, cost, and capabilities of Grok 4.1 Fast