xAI has unveiled its latest artificial intelligence model, Grok 4 Fast, describing it as a breakthrough in cost-efficient reasoning. Released in September 2025, the model is designed to deliver near-frontier performance while being significantly cheaper and faster than rivals.
What is Grok 4 Fast?
Grok 4 Fast builds on the company’s flagship Grok 4 model, but with a focus on speed and affordability. It features a 2 million token context window – allowing it to process very large documents or codebases – and a unified architecture that blends quick response and deep reasoning within a single system.
Unlike earlier systems, which used separate models for short answers versus extended reasoning, Grok 4 Fast adapts automatically, cutting response time and lowering costs.
How does it compare with other AIs?
On academic and reasoning benchmarks, Grok 4 Fast performs close to top models like OpenAI’s GPT-5 while surpassing earlier versions of xAI’s technology.
- On the AIME 2025 maths test, it scored 92%, comparable to GPT-5’s 94.6%.
- On LiveCodeBench, which measures coding ability, it achieved 80%, behind GPT-5 but ahead of Google’s Gemini 2.5 Pro and several other competitors.
- In search-related tasks, it outperformed rivals, ranking first in LMArena’s Search Arena, a competition testing real-world research skills.
The model is particularly strong in web-augmented queries. It can browse the internet and X (formerly Twitter), follow links, and analyse images and videos, giving it an edge in real-time information gathering.
Why “Fast”?
The “fast” label refers to more than speed of output. Grok 4 Fast uses about 40% fewer “thinking tokens” – the internal steps AI models take when reasoning – compared with Grok 4. This efficiency means it delivers answers more quickly while maintaining accuracy.
Independent reviewers at Artificial Analysis found that Grok 4 Fast is up to 47 times cheaper to run than comparable models, giving it what they call a state-of-the-art “price-to-intelligence ratio.”
Pricing and availability
For developers, Grok 4 Fast is available through the xAI API, OpenRouter, and Vercel AI Gateway. Two versions are offered:
- grok-4-fast-reasoning (for complex tasks)
- grok-4-fast-non-reasoning (for quick responses)
Pricing starts at $0.20 per million input tokens and $0.50 per million output tokens for shorter contexts, with discounts for cached input. For comparison, this is substantially lower than flagship models such as GPT-5 or Anthropic’s Claude Opus.
All xAI app users, including those on free plans, now have access to Grok 4 Fast – a move the company says will help “democratise advanced AI.”
Looking ahead
xAI says it will continue improving Grok 4 Fast with multimodal features, enabling the model to handle richer data like images, audio, and video.
While it may not always outperform the largest and slowest models on the toughest problems, Grok 4 Fast is aimed squarely at everyday use. For most tasks, its combination of speed, accuracy, and low cost could make it the model that people actually use.
