Even as OpenAI intensifies its legal battle with Elon Musk, the billionaire’s artificial intelligence firm, xAI, is pressing forward with the rollout of its flagship language model, Grok 3.
On Wednesday, xAI officially made Grok 3 and a smaller variant, Grok 3 Mini, available to developers through an application programming interface (API), marking the company’s most aggressive commercial push since its founding in 2023.
Grok 3, which was unveiled earlier this year as a competitor to OpenAI’s GPT-4o and Google’s Gemini, can analyze images and respond to complex queries. The model currently powers features on X, the social media platform formerly known as Twitter, which xAI acquired in March.
The new API offering allows developers to access Grok 3’s full capabilities, with pricing set at $3 per million input tokens and $15 per million output tokens — equivalent to roughly 750,000 input words and 750,000 generated words. The lighter Grok 3 Mini costs $0.30 and $0.50 per million input and output tokens respectively.
xAI is also offering speed-optimized versions of both models. Grok 3’s fast API tier will cost $5 per million input tokens and $25 per million output tokens, while Grok 3 Mini’s accelerated version is priced at $0.60 and $4 respectively.
The pricing positions Grok 3 on par with Anthropic’s Claude 3.7 Sonnet model, another high-performance generative AI offering with reasoning capabilities. However, Grok 3 remains more expensive than Google’s Gemini 2.5 Pro, which has generally outperformed it in benchmark testing. Industry analysts and critics have accused xAI of publishing selectively favorable benchmarks.
Adding to the controversy, developers on X have noted that the Grok 3 API supports a smaller context window — 131,072 tokens (around 97,500 words) — compared to the 1 million-token context that xAI previously claimed the model could handle.
Despite the model’s ambition and edgy branding, Grok has faced challenges in balancing its “unfiltered” identity with political neutrality. Earlier versions of the model were designed to provide answers that other chatbots might avoid, but studies have found that Grok often leaned left on social and political issues, including transgender rights and diversity initiatives.
Musk has blamed the perceived bias on training data sourced from public websites and has pledged to shift Grok closer to what he calls a “politically neutral” stance. It remains unclear whether Grok 3 has achieved this ideological recalibration, or how its responses might evolve as xAI scales up adoption.
The launch comes just days after OpenAI countersued Musk, accusing him of harassment and attempting to disrupt its corporate plans through legal and public pressure. Musk, who helped co-found OpenAI in 2015 before leaving the company, now views it as a key rival in the race to dominate the AI industry.
