xAI launches Grok 4 Fast with a 2M token context window

Elon Musk’s xAI has launched Grok 4 Fast, its newest artificial intelligence model that boasts high-level reasoning at significantly lower costs. xAI announced the launch of Grok 4 Fast in a blog post on Friday as its “latest advancement in cost-efficient reasoning models.” According to developers, it uses frontier-level intelligence with accessibility to benefit both enterprise and consumer users. The model features a two-million token context window and introduces a unified architecture that integrates reasoning and non-reasoning modes. This could help developers adjust the computational effort applied to their use cases without sacrificing performance. “2M context window!” CEO Elon Musk wrote on X, quoting the announcement for the new model. Grok 4 Fast has efficient tokens working on reduced prices xAI reported that Grok 4 Fast was trained end-to-end with reinforcement learning to autonomously decide when to use external resources such as code execution or web browsing. The tech company claimed it can move rapidly through links, analyze multimedia on X, and synthesize information in real time. In internal tests , the system required 40% fewer tokens on average than Grok 4 to complete complex reasoning tasks, a 98% reduction in the cost of reaching comparable benchmark results. Intelligence density benchmark. Source: xAI Independent evaluations from Artificial Analysis ranked Grok 4 Fast as having the best price-to-intelligence ratio among publicly available models. A chart released from the review showed Grok 4 Fast competing closely with systems such as GPT-5 , Gemini 2.5 Pro, and Claude 4.1 Opus, while being nearly 47 times cheaper to operate than Grok 4. Beyond synthetic benchmarks, Grok 4 Fast has been tested in LMArena, an independent leaderboard where models compete in text and search challenges. Under the “Menlo” codename, Grok 4 Fast topped the Search Arena with an Elo rating of 1,163. The margin was 17 points higher than its closest competitor, o3-search. For now, the model is available free of charge through OpenRouter and Vercel AI Gateway. xAI also launched the system in two variants: grok-4-fast-reasoning and grok-4-fast-non-reasoning. Both versions share the two-million token context window. Through the xAI API, pricing for grok-4-fast is set at $0.20 per million input tokens below 128,000 tokens, and $0.40 per million input tokens above that threshold. Output tokens cost $0.50 per million for smaller requests and $1.00 per million for larger ones. Cached input tokens are priced at $0.05 per million. xAI laid off AI tutors before Grok 4 Fast launch The launch comes just a week after xAI laid off at least 500 workers from its data annotation team, the largest unit within the company. These employees were responsible for labeling and categorizing data that trains the Grok models. Emails sent to staff on the evening of September 12 informed them of the layoffs, with those affected being promised continued salaries until November 30. However, their access to internal systems was cut off immediately. The AI company shared a post on X announcing it was recruiting specialist AI tutors, alongside plans to increase the size of its specialist tutor group from STEM fields by tenfold. The layoffs follow several high-profile exits from the company. Among them was Mike Liberatore, xAI’s chief financial officer, who recently stepped down. Musk is reportedly trying to reorient the company after his personal and political setbacks earlier this year. The 54-year-old billionaire has focused all his time on xAI since parting ways with President Donald Trump in June. Once aligned on technology and economic policy, and even sharing stages during Trump’s presidential campaign, the two clashed publicly, leading Musk to declare the government a “hopeless problem.” Get up to $30,050 in trading rewards when you join Bybit today