Grok 3 Mini

Overview

Grok 3 Mini is xAI's cost-efficient reasoning-capable model, released June 10, 2025. At $0.30/$0.50 per million tokens — roughly 90% cheaper than Grok 3 — it is one of the most affordable models from a major AI lab. Notably, independent benchmarks on the Artificial Analysis Intelligence Index score Grok 3 Mini (57) higher than Grok 3 full (45), making it not just cheaper but measurably better on certain tasks, particularly those involving reasoning.

This is a recurring theme in the current AI landscape: smaller, more focused models frequently outperform their larger counterparts on specific benchmarks. Grok 3 Mini appears to be xAI's best value offering and the recommended starting point for most API integrations.

Specifications

• Developer: xAI
• Model String: grok-3-mini
• Release Date: June 10, 2025
• Type: Large Language Model with reasoning capabilities
• Context Window: 131,072 tokens (131K)
• Knowledge Cutoff: November 2024 (with real-time X platform access)
• Access: xAI API, X (Twitter) Premium+
• Pricing: $0.30 per million input tokens / $0.50 per million output tokens

Capabilities

Reasoning at Low Cost: Includes reasoning capabilities — the ability to work through problems step-by-step before responding — at a price tier that significantly undercuts comparable reasoning models from OpenAI (o4-mini) and Anthropic.

Benchmark Performance: Scores 57 on the Artificial Analysis Intelligence Index vs. 45 for Grok 3 full — better performance at 90% lower cost. Particularly effective for tasks involving logical reasoning and structured analysis.

Real-Time X Access: Retains Grok 3's unique real-time X platform integration for current-events awareness.

Speed: Reasoning models typically have higher latency due to internal thinking steps; Grok 3 Mini balances this against its cost advantage.

Limitations

The 131K context window is the same constraint as Grok 3 — smaller than most competing models. Output pricing at $0.50 per million tokens is very low, but input pricing at $0.30 is comparable to Google's Gemini 2.5 Flash. For heavy output-generation tasks, this pricing structure is highly favorable; for input-heavy tasks (large document analysis), the context window limitation is the main constraint.

Recent Developments

• June 10, 2025 Launch: Released as xAI's cost-efficient alternative to Grok 3 full.
• Benchmark Surprise: The Mini outscoring the full model on the Artificial Analysis Intelligence Index attracted attention as a case study in how smaller reasoning-optimized models can outperform larger general-purpose ones.

Last Updated

February 26, 2026