xAI Launches Grok-Code-Fast-1: A High-Speed, Economical Agentic Coding Model

Elon Musk's Grok Has a New Coding Model That Emphasizes Speed

Grok is coming for the coding-focused customers of its bigger rivals. Elon Musk's xAI dropped a new AI model this week, specifically for developers, calling it a "speedy and economical reasoning model" meant to help with coding projects. The new addition, dubbed grok-code-fast-1, helps round out the family of Grok 4 models, released earlier this year. The coding model is agentic, meaning it is built to handle tasks without a ton of human oversight. xAI claims its model outperformed OpenAI's GPT-5, Anthropic's Claude 4 and xAI's own Grok 4, saying it could solve 70.8% of real-life software problems, as evaluated by the SWE-bench. Agentic AI tools have been a popular development in the ever-tightening AI race, with OpenAI, Anthropic and Microsoft each releasing a version of a coding agent. xAI's tool will be more similar to Microsoft's in that it's available via its API, i.e., not open to paying ChatGPT or Claude subscribers. You can access the new model for free for the next few days through a number of exclusive platforms, including Github Copilot, Cline, opencode, Cursor, Kilo Code, Roo Code and Windsurf. You'll need an API key, which is priced out by usage; you can learn more about that with this breakdown of Grok models. xAI also put together a prompt engineering guide specifically for the coding model to help developers get the most out of it. AI in coding projects can be a tricky conundrum. Generative AI tools like chatbots have proven to be effective tools to help beginners upskill quickly and help professionals problem-solve. Agentic AI is beginning to show how it can help automate nonessential tasks. That poses the question of whether automated tools could lead to widespread employment disruption, particularly for young software engineers. A new Stanford study finds a high risk for workers early in their careers with significant occupational exposure to AI, like software engineering and customer service. The regular Grok chatbot and X/Twitter assistant has had its share of controversies. Earlier this summer, the chatbot started producing extremely racist and antisemitic content, a few days after xAI founder Elon Musk posted that the company had "improved Grok significantly." xAI pulled the update soon after and shared that it had "taken action to ban hate speech before Grok posts on X." Bias is always an issue with AI chatbots, but Grok has illustrated how quickly the tech can spiral into being used to quickly produce potentially abusive content. (Disclosure: Ziff Davis, the parent company of CNET, in April filed a lawsuit against ChatGPT maker OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)

Grok's First Vibe-Coding Agent Has a High 'Dishonesty Rate'

Don't miss out on our latest stories. Add PCMag as a preferred source on Google. Elon Musk's xAI released its first agentic coding model, which claims to be "speedy and economical. " However, it also has "a higher dishonesty rate" than the company's flagship chatbot model, Grok 4. The AI startup designed the new model, grok-code-fast-1, specifically for coding tasks. It's free now for a limited time and accessible within GitHub Copilot, Cursor, Cline, Roo Code, Kilo Code, opencode, and Windsurf. "Grok-code-fast-1 has mastered the use of common tools like grep, terminal, and file editing, and thus should feel right at home in your favorite IDE," xAI says. But its propensity not to tell the truth could create problems for users. "We find that the dishonesty rate exceeds that of Grok 4," says the model card. The company attributes this in part to its "safety training, which teaches the model to answer all queries that do not express [a] clear intent to engage in specified prohibited activities." Translation: if it doesn't know the answer to your question, it might lie. If programmers ask the model if a certain part of the codebase is working, and it doesn't know, it may say "yes," when, in fact, the opposite is true. It might also confirm that it completed a test the engineer asked it to do when it did not. This could create blind spots and double work. It's not a major concern for xAI, which says it doesn't expect the model "to be widely used as a general-purpose assistant," like ChatGPT or the Grok chatbot. Vibe-coding agents are a new trend that stands to revolutionize the field, but they're far from perfect. One tool deleted a startup's entire client database on its own and deceived the user multiple times along the way. In fact, most of the large language models in the market today have behavioral issues, including blackmail, sabotage, lying, and telling the user what they want to hear (sycophancy). In a recent test, Anthropic and OpenAI examined each other's models and found these issues in almost all of them. Another eye-catching part of the Grok Code Fast 1 model card discusses the risk of someone using it to develop biological weapons. The company tested for this before release, along with issues related to cybersecurity and chemical knowledge. But bioweapons are the biggest risk, and "have the potential for the greatest scale of harm, [since] frontier models significantly lower the barrier to entry to the creation of bioweapons," xAI says. The results showed that Grok Code Fast 1 was worse than a human at "identifying issues in biological protocols," but it was better at "troubleshooting wet lab virology experiments." Again, xAI downplayed the issue, claiming that since the capabilities are similar to Grok 4, the new model "does not meaningfully change the risk landscape." Earlier this month, Anthropic updated the usage policy of its Claude chatbot to forbid using it to "synthesize, or otherwise develop, high-yield explosives or biological, chemical, radiological, or nuclear weapons or their precursors." Grok Code Fast 1 has secretly been out in the wild for the past week under the code name sonic. The xAI team says it "carefully monitored" feedback and deployed fixes, and plans to keep up a high rate of improvements "in days rather than weeks." At the same time, lying seems to be a particularly tough problem for AI companies to completely solve, at least in the short term.

Musk's xAI forays into agentic coding with new model

Aug 28 (Reuters) - Elon Musk's artificial intelligence startup, xAI, on Thursday released a new "speedy and economical" agentic coding model, marking its entry into a key focus area for AI companies. An agentic coding tool is an AI-powered software application that can autonomously perform coding-related tasks. XAI's model, called grok-code-fast-1, would be available for free for a limited time, with select launch partners including GitHub Copilot and Windsurf. Its "strength lies in delivering strong performance in an economical, compact form factor, making it a versatile choice for tackling common coding tasks quickly and cost-effectively," xAI said. AI companies such as OpenAI and Microsoft (MSFT.O), opens new tab are focusing on making AI coding assistants available to its users. Microsoft in May introduced the GitHub Copilot feature, a coding agent, at its annual Build software developer conference. CEO Satya Nadella had in April said that 20% to 30% of overall code in Microsoft is being written by AI. ChatGPT maker OpenAI's agent, called Codex, was made available to ChatGPT Plus users in June. XAI sued Apple (AAPL.O), opens new tab and OpenAI in U.S. federal court in Texas on Monday, accusing them of illegally conspiring to thwart competition for AI. Reporting by Juby Babu in Mexico City; Editing by Maju Samuel Our Standards: The Thomson Reuters Trust Principles., opens new tab

Elon Musk's xAI releases agentic coding model aimed at high speed and affordability - SiliconANGLE

Elon Musk's xAI releases agentic coding model aimed at high speed and affordability Elon Musk's xAI Holdings Corp. has released grok-code-fast-1, a dedicated agentic coding artificial intelligence model that is extremely speedy and designed to strike a "compelling balance between performance and cost." In a market that is quickly becoming cluttered with models offering coding capabilities, xAI said Thursday it built the model from the ground up and built it with a brand-new architecture fit for task. "Throughout the training process, we collaborated closely with our launch partners to refine and sharpen the model's behavior inside their agentic platforms," the company said. "Grok-code-fast-1 has mastered the use of common tools like grep, terminal, and file editing, and thus should feel right at home in your favorite IDE." At launch, the new model will be available for free for a limited time within popular AI-enabled coding platforms, including GitHub Copilot, Cursor, Cline, Roo Code, Kilo Cline, Opencode and Windsurf. The model supports function calling, structured outputs and reasoning with a 256,000-token context window. This window size enables the model to recall the equivalent of hundreds of pages of text or code simultaneously, allowing it to efficiently review large portions of codebases while working. As for speed, according to xAI's own benchmarks, the new model can execute at about 160 tokens per second. Compared with other popular models on the market on the same xAI released benchmarks, OpenAI's GPT-5 averages about 50.1 tokens per second, Gemini 2.5 Pro hit about 92.4 and Claude 4 Sonnet reached 78.7. "In early testing, Grok Code Fast has shown both its speed and quality in agentic coding tasks," said Mario Rodriguez, chief product officer of GitHub Inc. "Empowering developers with powerful tools is a core part of our mission at GitHub Copilot, and this is a compelling new option for our developers." Last week, xAI stealthily released grok-code-fast-1 under the codename sonic. During this phase, the research team monitored community channels and adjusted the model according to feedback. The company said on a full subset of SWE-Bench-Verified, a human-validated evaluation of the AI model's ability to solve real-world software engineering problems, the model received a 70.8% using an internal system. In comparison, GPT-5 received a 74.9% (with thinking) and Claude Sonnet 4 achieved 72.7%. The model was designed to be flexible across various coding languages, with high proficiency in TypeScript, Python, Java, Rust, C++ and Go. The model is available through the xAI application programming interface service for developers, priced at 20 cents per 1 million input tokens, $1.50 per 1 million output tokens, and two cents per 1 million cached input tokens. Given that xAI will be competing against multiple other models on the market that provide coding capabilities, the company said it intends to deliver consistent updates and improvements on the order of days rather than weeks. The company teased a new variant that supports multimodal inputs, parallel tool calling and extended context as the first upcoming release.

xAI releases agentic coding model aimed at high speed and affordability - SiliconANGLE

xAI releases agentic coding model aimed at high speed and affordability Elon Musk's xAI Holdings Corp. released grok-code-fast-1 on Thursday, a dedicated agentic coding artificial intelligence model that is extremely speedy and designed to strike a "compelling balance between performance and cost." In a market that is quickly becoming cluttered with models offering coding capabilities, xAI said it built the model from the ground up and built it with a brand-new architecture fit for task. "Throughout the training process, we collaborated closely with our launch partners to refine and sharpen the model's behavior inside their agentic platforms," the company said. "Grok-code-fast-1 has mastered the use of common tools like grep, terminal, and file editing, and thus should feel right at home in your favorite IDE." At launch, the new model will be available for free for a limited time within popular AI-enabled coding platforms, including GitHub Copilot, Cursor, Cline, Roo Code, Kilo Cline, Opencode and Windsurf. The model supports function calling, structured outputs and reasoning with a 256,000 token context window. This window size enables the model to recall the equivalent of hundreds of pages of text or code simultaneously, allowing it to efficiently review large portions of codebases while working. As for speed, according to xAI's own benchmarks, the new model can execute at around 160 tokens per second. Compared to other popular models on the market in the same xAI released benchmarks, OpenAI's GPT-5 averages around 50.1 tp/s, Gemini 2.5 Pro hit around 92.4 tp/s and Claude 4 Sonnet reached 78.7 tp/s. "In early testing, Grok Code Fast has shown both its speed and quality in agentic coding tasks," said Mario Rodriguez, chief product officer of GitHub Inc. "Empowering developers with powerful tools is a core part of our mission at GitHub Copilot, and this is a compelling new option for our developers." Last week, xAI stealthily released grok-code-fast-1 under the codename sonic. During this phase, the research team monitored community channels and adjusted the model according to feedback. The company said on a full subset of SWE-Bench-Verified, a human-validated evaluation of the AI model's ability to solve real-world software engineering problems, the model received a 70.8% using an internal system. In comparison, GPT-5 received a 74.9% (with thinking) and Claude Sonnet 4 achieved 72.7%. The model was designed to be flexible across various coding languages, with high proficiency in TypeScript, Python, Java, Rust, C++, and Go. The model is available through the xAI application protocol interface service for developers, priced at $0.20 per 1 million input tokens, $1.50 per 1 million output tokens, and $0.02 per 1 million cached input tokens. Given that xAI will be competing against multiple other models on the market that provide coding capabilities, the company said it intends to deliver consistent updates and improvements on the order of days rather than weeks. The company teased a new variant that supports multimodal inputs, parallel tool calling and extended context as the first upcoming release.

You Can Now Try xAI's GrokCode Fast 1 Model With Agentic Coding for a Week

Grok Code Fast 1, a new lightweight coding-focused artificial intelligence (AI) model, was released by xAI on Thursday. The Elon Musk-owned AI startup highlighted that this model comes with agentic capabilities and is designed as a full-stack coding agent that can be used to build both the backend and the front-end of a software. The company has touted its responsiveness, and the company claims it can generate output faster than coding tools offered by rivals. For a week, the model is available without any usage cost on popular third-party platforms. Grok Code Fast 1 Is Built on a New Architecture In a newsroom post, xAI detailed its latest AI coding model. The company said that existing coding models often become slow when executing functions requiring reasoning loops and tool calling. Solving the pain point, Grok Code Fast 1 is said to offer a faster output generation on agentic workflows. While the company did not share the details, it claimed that the new coding model is built on a new architecture, which was paired with programming-related content for pre-training and real-world pull requests for post-training. The company said it also partnered with several third-party platforms to refine the model behaviour while carrying out agentic tasks. Notably, Grok Code Fast 1 is claimed to be adept at calling tools such as grep, terminal, and file editing. xAI claimed this interactive game front-end was built entirely by Grok Code Fast 1 (on Cursor) Photo Credit: xAI As per the company, the AI model can perform tasks across the full software development stack, and it supports TypeScript, Python, Java, Rust, C++, and Go coding languages. "It can complete common programming tasks with minimal oversight," the post claimed. Based on internal evaluations, xAI said that the model achieved a 70.8 percent score on the full subset of SWE-Bench-Verified benchmark. For the next seven days, starting from August 28, xAI is offering Grok Code Fast 1 for free on the platforms of its launch partners. This means developers can use the model without incurring any charges on GitHub Copilot, Cursor, Cline, Kilo Code, Roo Code, opencode, and Windsurf. The company did not mention whether it is applying any rate limits during this period. After the limited free period, the AI coding model will be available at the price of $0.20 (roughly Rs. 17) per million input and $1.50 (roughly Rs. 132) per million output tokens. For cached input tokens, the model will charge $0.02 (roughly Rs. 1.7).

Musk's xAI forays into agentic coding with new model - The Economic Times

XAI's model, called grok-code-fast-1, would be available for free for a limited time, with select launch partners including GitHub Copilot and Windsurf.Elon Musk's artificial intelligence startup, xAI, on Thursday released a new "speedy and economical" agentic coding model, marking its entry into a key focus area for AI companies. An agentic coding tool is an AI-powered software application that can autonomously perform coding-related tasks. XAI's model, called grok-code-fast-1, would be available for free for a limited time, with select launch partners including GitHub Copilot and Windsurf. Its "strength lies in delivering strong performance in an economical, compact form factor, making it a versatile choice for tackling common coding tasks quickly and cost-effectively," xAI said. AI companies such as OpenAI and Microsoft are focusing on making AI coding assistants available to its users. Microsoft in May introduced the GitHub Copilot feature, a coding agent, at its annual Build software developer conference. CEO Satya Nadella had in April said that 20% to 30% of overall code in Microsoft is being written by AI. ChatGPT maker OpenAI's agent, called Codex, was made available to ChatGPT Plus users in June. XAI sued Apple and OpenAI in U.S. federal court in Texas on Monday, accusing them of illegally conspiring to thwart competition for AI.

Musk's xAI forays into agentic coding with new model

(Reuters) -Elon Musk's artificial intelligence startup, xAI, on Thursday released a new "speedy and economical" agentic coding model, marking its entry into a key focus area for AI companies. An agentic coding tool is an AI-powered software application that can autonomously perform coding-related tasks. XAI's model, called grok-code-fast-1, would be available for free for a limited time, with select launch partners including GitHub Copilot and Windsurf. Its "strength lies in delivering strong performance in an economical, compact form factor, making it a versatile choice for tackling common coding tasks quickly and cost-effectively," xAI said. AI companies such as OpenAI and Microsoft are focusing on making AI coding assistants available to its users. Microsoft in May introduced the GitHub Copilot feature, a coding agent, at its annual Build software developer conference. CEO Satya Nadella had in April said that 20% to 30% of overall code in Microsoft is being written by AI. ChatGPT maker OpenAI's agent, called Codex, was made available to ChatGPT Plus users in June. XAI sued Apple and OpenAI in U.S. federal court in Texas on Monday, accusing them of illegally conspiring to thwart competition for AI. (Reporting by Juby Babu in Mexico City; Editing by Maju Samuel)

How xAI's new coding model works and what it means for developers

Developers gain powerful AI coding tool as xAI enters competitive market xAI, the artificial intelligence company founded by Elon Musk, has recently unveiled grok-code-fast-1, a groundbreaking agentic coding model designed to revolutionize how developers approach software development. Launched on August 28, 2025, this model is engineered to deliver high performance in a compact, cost-effective form, making it a game-changer for coders tackling everyday programming tasks. This feature explores how grok-code-fast-1 works, its unique capabilities, and its implications for developers across the globe. Also read: The complete LLM showdown: Testing 5 major AI models for real-world performance Unlike general-purpose large language models, grok-code-fast-1 was specifically designed for agentic coding, AI-driven software that autonomously performs coding tasks with minimal oversight. xAI crafted this model from scratch, using a new architecture optimized for speed and efficiency. The model's pre-training corpus is rich with programming content, ensuring it understands the nuances of various programming languages and development workflows. The model excels in languages like TypeScript, Python, Java, Rust, C++, and Go, making it versatile across the full software development stack. It can handle tasks ranging from building projects from scratch to performing precise bug fixes and answering complex codebase queries. Its strength lies in its ability to deliver rapid, high-quality responses, allowing developers to iterate quickly without bottlenecks. Grok-code-fast-1 is described as "ridiculously fast," potentially the fastest coding model available today. Developers using tools like Cursor or Cline have reported that its quick response times enable a workflow shift toward smaller, focused tasks rather than large, monolithic prompts. This iterative approach allows coders to break down complex features into manageable phases, refining outputs with minimal latency. For example, a developer building a battle simulator prototype completed it in less than a day by leveraging grok-code-fast-1's rapid iterations. xAI's holistic evaluation approach combines public benchmarks with real-world testing. On the SWE-Bench-Verified subset, grok-code-fast-1 scored an impressive 70.8% using xAI's internal harness, showcasing its capability in realistic software engineering scenarios. While benchmarks like SWE-Bench provide valuable insights, xAI emphasizes that grok-code-fast-1's true strength is in its end-user experience, particularly in agentic coding workflows. Also read: Launching tomorrow, today: How Launch.today is reimagining web building with AI The model is accessible via the xAI API, with a pricing structure of $0.20 per million input tokens, $1.50 per million output tokens, and $0.02 per million cached input tokens. For a limited time, it's available for free with launch partners like GitHub Copilot and Windsurf, making it easy for developers to test its capabilities. It integrates seamlessly with tools like Cline's Plan/Act modes and MCP tools, requiring no special configuration. xAI provides a Prompt Engineering Guide to help developers maximize grok-code-fast-1's potential. The guide recommends breaking tasks into smaller, focused prompts to leverage the model's speed and steerability. This approach ensures precise outputs and allows developers to maintain control over complex projects. Grok-code-fast-1's speed and versatility enable developers to work more efficiently. By automating routine coding tasks, such as writing boilerplate code, debugging, or generating multi-language snippets, it frees up time for creative problem-solving. Its ability to handle zero-to-one projects and surgical bug fixes makes it a valuable tool for both solo developers and large teams. The model's economical design and temporary free access lower the barrier to entry for developers, especially those in resource-constrained environments. Its compatibility with popular development tools like GitHub Copilot and Cline ensures that developers can integrate it into existing workflows without significant overhead. This democratization aligns with xAI's mission to empower developers with powerful, accessible tools. grok-code-fast-1 enters a competitive landscape dominated by models like OpenAI's Codex and Microsoft's GitHub Copilot. Microsoft's CEO Satya Nadella noted that 20-30% of Microsoft's code is already AI-written, highlighting the growing reliance on coding assistants. grok-code-fast-1's focus on speed, cost-effectiveness, and real-world performance positions it as a compelling alternative, particularly for developers prioritizing rapid iteration and cost savings. While grok-code-fast-1 offers significant advantages, its high steerability requires developers to implement their own safety filters, which could be a double-edged sword. Without careful prompt engineering, there's a risk of generating incorrect or suboptimal code. Additionally, as with other AI coding tools, developers must validate outputs to ensure reliability, especially for mission-critical applications. xAI's roadmap includes expanding API capabilities, such as adding code execution and agentic capabilities that allow Grok to run tools or scripts autonomously. These features could further enhance grok-code-fast-1's utility, enabling it to handle more complex workflows, such as automated testing or deployment. The model's integration with xAI's Colossus supercomputer, which plans to scale to 200,000 GPUs by late 2025, suggests that future iterations will push the boundaries of AI-driven development. xAI's grok-code-fast-1 represents a significant leap forward in agentic coding, offering developers a fast, economical, and versatile tool to tackle modern software development challenges. Its purpose-built architecture, seamless integration, and focus on real-world performance make it a standout in the AI coding landscape. For developers, it promises enhanced productivity, greater accessibility, and the potential to redefine how code is written and iterated. As xAI continues to innovate, grok-code-fast-1 is poised to become an essential ally for coders navigating the fast-evolving world of software development.