Grok 4 Fast

2025-09-19 · Publisher : xAI · Domain : Artificial Intelligence

xAI launches Grok 4 Fast, a cost-efficient reasoning model with a 2 million token context window and a unified reasoning and non-reasoning architecture.

xAI launches Grok 4 Fast, a cost-efficient reasoning model with a 2 million token context window and a unified reasoning and non-reasoning architecture. ### What changes - Performance close to Grok 4 with 40% fewer thinking tokens on average, hence a 98% cost reduction for the same result, according to xAI. - A single model for both reasoning and non-reasoning modes. - Native web and X search, with tool use trained through reinforcement learning. - Available to all users of grok.com and the apps, including free users. ### Specifications - Context: 2 million tokens. - API identifiers: grok-4-fast-reasoning and grok-4-fast-non-reasoning. - Availability: Grok apps, xAI API, OpenRouter, Vercel AI Gateway. - Knowledge cutoff: not disclosed in the post. ### Pricing Per million tokens, in USD (xAI API). | | Input | Output | Cached input | |---|---|---|---| | Under 128,000 tokens | $0.20 | $0.50 | $0.05 | | 128,000 tokens and above | $0.40 | $1.00 | $0.05 | ### Benchmarks Scores published by the vendor (pass@1, without tools unless stated otherwise). | Benchmark | Grok 4 Fast | Grok 4 | |---|---|---| | GPQA Diamond | 85.7% | 87.5% | | AIME 2025 | 92.0% | 91.7% | | HMMT 2025 | 93.3% | 90.0% | | Humanity's Last Exam | 20.0% | 25.4% | | LiveCodeBench (Jan.-May) | 80.0% | 79.0% | | BrowseComp | 44.9% | 43.0% | Grok 4 Fast ranks first on LMArena's Search Arena (1163 Elo). ### Key takeaways A model close to Grok 4 on many tasks for a fraction of the price, with a very long context.

Tags : LLM, xAI, Grok, Grok 4 Fast, Raisonnement