Grok 3 and Grok 3-mini API available: 50 cents per output mtok

X has made Grok 3 API-accessible and launched a smaller sibling, Grok 3-mini, the company announced on April 19, 2025. Grok 3 had been available for a couple of months but lacked an API option, and the new release adds programmatic access alongside a low-cost mini model.

What changed

Grok 3 is now accessible via API, and X published developer documentation at docs.x.ai/docs/overview. The company also introduced Grok 3-mini, described as a “baby brother” to Grok 3. The mini model is priced at 50 cents per output mtok, and it explicitly "claims to be competitive with much larger frontier models," while offering full reasoning traces.

The primary functional change is API availability for a model that had previously been available but not exposed as a service endpoint, enabling developers and integrators to call Grok 3 programmatically. The addition of Grok 3-mini introduces a lower-cost option with a stated design emphasis on exposing internal reasoning steps alongside model outputs.

How Grok 3-mini positions itself

Grok 3-mini is presented as a cost-conscious alternative to larger frontier models, with two concrete selling points in the announcement: a published unit price of 50 cents per output mtok, and the feature of full reasoning traces. The source text frames Grok 3-mini as competitive with larger models on a cost and capability basis, by claiming parity with larger frontier models while providing transparent internal reasoning.

Developers interested in experimenting or integrating Grok 3 and Grok 3-mini are pointed to the X developer docs URL. The API availability converts prior access patterns, which were limited to non-API interaction, into a model-as-a-service workflow that supports direct programmatic usage.

Why it matters

API availability converts a model from a demo or limited-release experience into a tool that product teams, researchers, and hobbyist builders can integrate into applications. The presence of a lower-priced mini variant priced at 50 cents per output mtok reduces the per-call cost barrier, which may drive more trial usage and prototype development. The explicit offering of full reasoning traces from Grok 3-mini responds to growing interest in model transparency and explainability, potentially making the outputs easier to inspect or debug in agentic or high-assurance workflows.

Moreover, positioning Grok 3-mini as competitive with larger frontier models at a lower cost signals a continued market trend: vendors offering smaller, cheaper models that aim to match selective capabilities of much larger systems while adding differentiating features like trace outputs.

What to watch

Watch for developer uptake and real-world cost comparisons between Grok 3-mini and larger models, particularly in integrations that depend on model reasoning traces. Also monitor updates to the docs.x.ai developer documentation for API details, rate limits, and example workflows that will show how X intends customers to use the new endpoints.