Grok 4 Fast arrives as cheaper alternative to xAI's Grok-4
What's the story
Elon Musk's artificial intelligence (AI) company, xAI, has unveiled a new model called Grok 4 Fast. The innovative system is a more affordable version of the existing Grok 4. It has been designed for both enterprise and consumer use cases. The unique selling point of this model is its ability to perform reasoning and non-reasoning tasks in a single framework, making it highly efficient.
Cost efficiency
Grok 4 Fast offers a better cost-to-performance ratio
Grok 4 Fast is a more cost-effective alternative to its predecessor, Grok 4. The new model uses an average of 40% fewer "thinking tokens" than the previous version, while still maintaining similar accuracy across benchmarks. According to Artificial Analysis's independent review, Grok 4 Fast has a lower cost-to-performance ratio and can achieve results similar to those of Grok 4 at up to a whopping 98% less spending.
Benchmark results
Model's performance on various benchmarks
Grok 4 Fast has shown impressive performance on various benchmarks. It scored 85.7% on GPQA Diamond, 92% on AIME 2025, and 93.3% on HMMT 2025 - all of which are pretty close to what Grok 4 achieved. The model also improved in code execution and search-based benchmarks with a pass rate of up to an impressive 95% on SimpleQA and a decent score of about 74% on X Bench Deepsearch.
Model features
Key features of Grok 4 Fast
Grok 4 Fast boasts a massive two million token context window, allowing it to process large inputs. It was trained with reinforcement learning to make it more efficient. Unlike its predecessors that required separate reasoning and non-reasoning models, Grok 4 Fast combines both into one system. This integration cuts down latency and cost significantly. The model is available on xAI's website, mobile apps, OpenRouter, and Vercel AI Gateway with free access on some platforms during launch period.
Pricing details
Pricing and availability
Grok 4 Fast comes in two versions - grok-4-fast-reasoning and grok-4-fast-non-reasoning. Both support the two million token limit. The pricing starts at $0.20 per million input tokens for smaller workloads, increasing with token usage. This makes it an ideal choice for large projects or anyone looking for a fast, affordable AI solution without compromising on accuracy.