xAI launches an API for Grok 3 and Grok 3 Mini with “reasoning” capabilities; Grok 3 costs $3/$15 per million input/output tokens, Grok 3 Mini costs $0.30/$0.50
Context & Ripple Effects
xAI’s path moved from a Grok-2 beta whose API availability was planned to a Grok-3 beta for X Premium+ users. This release turns the newer Grok 3 and smaller Grok 3 Mini models into products developers can invoke programmatically.
The split between a higher-priced Grok 3 and far cheaper Grok 3 Mini makes the launch as much a packaging and usage-pricing decision as a model release. It creates a clear cost/performance choice for application builders.
First-order effects
- Developers can now integrate Grok 3 and Grok 3 Mini through an API rather than relying on access through X subscriptions; Grok 3 is priced at $3 per million input tokens and $15 per million output tokens.
- The $0.30/$0.50-per-million-token pricing for Grok 3 Mini gives customers a substantially lower-cost option for workloads that do not require the flagship model.
Second-order effects
- Teams adopting Grok must route tasks between the two models based on output quality and token costs, making workload design central to the effective price of reasoning features.
- The published rates give buyers a direct benchmark for evaluating competing reasoning-model APIs and increase pressure on providers to differentiate on both capability and inference cost.
Third-order effects
- If providers continue pairing reasoning models with sharply differentiated API tiers, model selection will increasingly be an application-level unit-economics decision rather than a single-vendor subscription choice.
- This points toward reasoning capability becoming commercially standardized through metered APIs, with durable advantage depending on the cost of useful outputs rather than model access alone.
The trend: Reasoning models are shifting from subscriber-facing launches into tiered, metered API products where developers optimize capability against token economics.