xAI Launches Grok4Fast
xAI has introduced Grok4Fast, a lightweight flagship model that the company claims delivers performance comparable to Grok4 while reducing computation by 40%. According to AIbase, this significant efficiency boost cuts task costs by up to 98%.
Balancing Performance and Efficiency
Grok4Fast has shown outstanding results in multiple benchmarks—for example, scoring 85.7% on GPQA Diamond and 92.0% on AIME2025, achievements on par with Grok4 and even GPT-5. xAI highlighted that the model achieves this by reducing “thinking tokens,” using on average 40% fewer tokens than Grok4 to produce similar results. This efficiency advantage is particularly notable when handling tasks that require complex reasoning.
Integrated Architecture and External Tools
Unlike earlier versions that relied on separate models for different tasks, Grok4Fast integrates both approaches into a single architecture, with system prompts controlling behavior—reflecting the latest trend in hybrid models.
The model also demonstrates strong external tool capabilities, including web browsing and code execution. On benchmarks such as BrowseComp and X Bench Deepsearch, Grok4Fast outperformed Grok4. In the LMArena-Search benchmark, it even surpassed the previously leading OpenAI o3-websearch model. On Text Arena, Grok4Fast currently ranks eighth, ahead of other models of similar scale.
Availability and Pricing
Grok4Fast is available in two versions: one optimized for reasoning-intensive tasks, and another designed for quick answers. Both versions support a 2-million-token context window. The model can be accessed via grok.com, iOS and Android apps, and the xAI API. Pricing ranges from $0.05 to $1.00 per million tokens, depending on token type. For now, users can also try Grok4Fast for free via OpenRouter and Vercel.
© Copyright notes
The copyright of the article belongs to the author, please do not reprint without permission.
