Grok 3 Mini: xAI's budget model

Grok 3 Mini is xAI's budget option — the previous-generation Mini that still earns a spot in the lineup because it's cheap, fast, and good enough for routine work. It ships with a 131K-token context window, native X / web / news search (xAI's signature feature), and prices at $0.30 per million input tokens and $0.50 per million output. The case for Grok 3 Mini: quick chat responses, lightweight summarization, classification, and routine agent steps where you want xAI's fresh web grounding at a budget price. The case against: for most new workloads Grok 4.1 Fast is the obvious pick — it's slightly cheaper on input, has a 15x larger context window, supports vision, and reasons better. Grok 3 Mini exists mostly for callers who specifically benchmarked against it and want a stable, identical SKU to keep their pipelines reproducible. New projects should default to Grok 4.1 Fast unless there's a reason not to.

Specs

Best at

• Quick chat responses — low latency for everyday Q&A.
• Routine classification & summarization — accurate enough at very low cost.
• X-grounded research — native X/web/news search at budget pricing.
• Reproducible pipelines — stable older SKU for benchmarks and regression tests.

Where it loses

• Long context — 131K is small next to Grok 4.1 Fast's 2M.
• Vision — text only; cannot read screenshots or charts.
• Hard reasoning — any of the 4.x Grok models reason noticeably better.
• Coding — pick Sonnet 4.6 or Opus 4.7 for refactor work.

Should you use it?

• If you have an existing pipeline built on Grok 3 Mini — keep it; it still works.
• For new projects, default to Grok 4.1 Fast instead. Same provider, lower input price, 15x context, vision, better reasoning.

Frequently asked questions