Elon Musk’s AI company, xAI, has launched an API for its Grok 3 model, offering developers access to its latest artificial intelligence system.Â
This comes despite the ongoing legal challenges Musk is facing, including a countersuit from OpenAI.
The Grok 3 model, introduced several months ago, serves as xAI’s competitor to AI models such as OpenAI’s GPT-4o and Google’s Gemini.
Grok 3 has the ability to analyze images and answer questions, and it is integrated into several features of Musk’s social media platform, X, which xAI acquired in March.
Through its API, xAI provides two versions of Grok 3: the full Grok 3 and Grok 3 Mini, which also includes reasoning capabilities.
The pricing is set at $3 per million tokens for input and $15 per million tokens for output on the standard Grok 3 model.
The Mini version is priced at a lower rate of $0.30 per million tokens for input and $0.50 per million tokens for output.
For faster processing speeds, premium versions of both models are available, with Grok 3 charging $5 per million input tokens and $25 per million output tokens, while Grok 3 Mini is priced at $0.60 per million input tokens and $4 per million output tokens.
Compared to competitors, Grok 3’s pricing is on par with Anthropic’s Claude 3.7 Sonnet, which also includes reasoning capabilities.
 However, it is priced higher than Google’s Gemini 2.5 Pro, which has demonstrated superior performance in several benchmarks.
Some users on X have noted that Grok 3’s API has a smaller context window than initially promised, supporting only up to 131,072 tokens (around 97,500 words), well below the 1 million tokens xAI had mentioned in earlier reports.
When Grok was first introduced, Musk described it as a more direct, unfiltered AI model, aiming to provide responses that might be avoided by other systems.
This approach was seen in the earlier versions, where Grok was willing to give more controversial and uncensored answers.
However, previous models also showed some political biases, leaning left on certain topics.
Musk attributed these biases to Grok’s training data and has since stated his intention to make Grok more politically neutral.
Whether the Grok 3 model has fully achieved this remains unclear, but it continues to be a focus of development for xAI.
