Elon Musk-owned xAI has unveiled Grok 4 Fast, a cost-effective AI model designed for both enterprise and consumer use. The model combines reasoning and non-reasoning capabilities in a single framework, eliminating the need for separate models and reducing latency and costs.

According to xAI, Grok 4 Fast uses 40% fewer “thinking tokens” than the previous Grok 4 while maintaining comparable accuracy across benchmarks. Independent reviews confirm its lower cost-to-performance ratio, requiring up to 98% less spending to achieve results similar to Grok 4.

Benchmark tests show Grok 4 Fast scoring 85.7% on GPQA Diamond, 92% on AIME 2025, and 93.3% on HMMT 2025. The model also improved code execution and search-based benchmarks, including a 95% pass rate on SimpleQA.

With a 2 million token context window, Grok 4 Fast can handle large inputs efficiently. It is trained using reinforcement learning, making it faster, smarter, and cheaper. Two versions are available-grok-4-fast-reasoning and grok-4-fast-non-reasoning, starting at $0.20 per million tokens, with free access on OpenRouter and Vercel for a limited period.

Grok 4 Fast is accessible on grok.com, iOS, Android, and via xAI API, making it a versatile AI solution for developers and businesses alike.