Grok 3 is now available via API to compete with OpenAI and Google

xAI, Elon Musk’s AI company and a direct competitor to OpenAI, Anthropic (creator of Claude chatbot), Google and all other major players, released its most advanced model, Grok 3, almost two months ago. Until now, the model was accessible only through the Grok website, X platform and Grok mobile apps. Yesterday, they released an API for letting developers build AI-driven applications and solutions with Grok 3.

The API is available in two different variants. “Grok 3” and “Grok 3 Mini (with Thinking)”.

Grok 3 is xAI’s flagship model and a strong rival for models like Gemini 2.5 Pro, OpenAI’s GPT-4o, Claude 3.7 Sonnet, and DeepSeek-R1. It excels at data extraction, coding, and text summarization. xAI has implanted deep domain knowledge in finance, healthcare, law, and science into the core of this model.

Grok 3 Mini is on the other hands a lightweight model that is able to “think” before responding. It’s fast, and great for logic-based reasoning tasks that do not require deep domain knowledge. Its raw thinking process is also accessible through the API.

Each of these two variants also comes in two speeds; standard and fast. The exact model names are as follows:

• grok-3

• grok-3-fast

• grok-3-mini

• grok-3-mini-fast

According to xAI, the faster models are identical to their respective standard version and they point to the same underlying model. The fast versions are suggested to be used for latency-sensitive applications, and the normal ones for reduced cost.

One important detail that should be pointed out is that the larger main model “grok-3” does not come with reasoning capabilities. However, the mini version of the grok-3 provides reasoning. Honestly, We have to admit that it is a bit confusing, but that is how it currently stands.

There's a significant price difference between the Grok 3 models, provided by this new API. The standard Grok 3 charges $3 for every million input tokens (around 750,000 words) and $15 for every million output tokens. In contrast, the smaller Grok 3 Mini offers much lower rates: just $0.30 per million input tokens and $0.50 per million output tokens.

Grok 3’s API is a bit pricier than some of its competitors. For instance, Google's recently-released Gemini 2.5 Pro is priced at $1.25 per million input tokens and $10 per million output tokens for prompts up to a massive 200,000 tokens (over 750,000 words). For even larger prompts, a capability many competitors lack, the cost increases to $2.50 per million input tokens and $15 per million output tokens.

The context window of all of these new Grok 3 API models are set to only 131,072 tokens. The context window or “context length” of a large language model (LLM) is the amount of text, measured in tokens, that the model can consider or remember in a single prompt. Therefore, a larger context window is always better. xAI had previously promised to launch an API with a context window as large as 1M tokens, so it’s a bit disappointing as this API can take up to only 1/10th of the input size of what it was supposed to be able to process.

Unfortunately, unlike Grok on grok.com or on X platform, the Grok models on this new API are not connected to the internet; therefore they have no access to real-time information. The knowledge cutoff date for the Grok 3 model family is November 17, 2024 and Grok has no knowledge of current events or data beyond what was present in its training data.

One of the great capabilities of Grok 3 is that it is a multi-modal AI model, which means it can process both text and images. The Grok 3 API can handle images of up to 10MB in size that are in JPG/JPEG or PNG format. Luckily, there is no limit to the number of images that you can feed into the model.

For more details, benchmarks and tutorials visit xAI’s documentation.