xAI drops Grok 3 Mini to fuel the reasoning price war

The race to the bottom in AI inference pricing is officially out of control.

Following Google's recent cost cuts with Gemini 2.5 Flash, xAI is turning up the pressure. The company just released Grok 3 Mini and made both Grok 3 and its lightweight sibling available to developers via the xAI API.

Here is what Elon Musk's AI venture actually shipped.

Pick your speed, pick your thinking budget

The Grok 3 lineup now features six different configurations: flagship Grok 3, Grok 3 Fast, and four distinct flavors of Grok 3 Mini that mix fast or standard speeds with either low or high reasoning budgets.

Unlike the flagship Grok 3, which handles complex tasks without an explicit thinking step, Grok 3 Mini was engineered from scratch around an integrated reasoning process.

According to early benchmarks from independent tracking firm Artificial Analysis, Grok 3 Mini Reasoning with high compute actually outperformed heavyweights like DeepSeek R1 and Claude 3.7 Sonnet (configured with a 64k reasoning budget) on its combined Intelligence Index.

The pricing is where xAI is really swinging. Standard Grok 3 Mini costs just $0.30 per million input tokens and $0.50 per million output tokens—nearly an order of magnitude cheaper than OpenAI's o4-mini or Google's Gemini 2.5 Pro. A high-speed endpoint is also available at $0.60 input and $4.00 output.

There is an obvious tradeoff in latency, though. On standard endpoints, flagship Grok 3 spits out 500 tokens in roughly 9.5 seconds, while the reasoning Mini variant takes 27.4 seconds to grind through the same length.

xAI is also returning full reasoning traces in every API response so developers can inspect the model's intermediate logic, though research shows these machine thoughts can still be misleading.

Why it matters

We think xAI just made frontier-grade reasoning accessible enough to become an impulse buy for developers.

At thirty cents a million input tokens, solo builders and early-stage startups can afford to pipe complex math, automated code auditing, and multi-step logic into their apps without sweating monthly server bills. That said, benchmark dominance rarely translates one-to-one into messy real-world tasks, and waiting nearly half a minute for a response means you should test this on offline workflows before betting customer-facing products on it.

For the broader industry, this confirms xAI can compete toe-to-toe with Google and OpenAI on both performance and price. By landing both Grok 3 and Grok 3 Mini in the top five of Artificial Analysis rankings, xAI is showing that cost-efficient reasoning is no longer a DeepSeek monopoly, forcing legacy labs to rethink their high API margins.