Flow - 4x faster than typing
Mini toolsGrok 3 Think
By xAI
Model family: Grok
Grok 3 Think is a selectable reasoning mode in the Grok 3 family designed for complex problem-solving: it engages longer “deliberation,” corrects itself, and can show a raw reasoning trace via the API for auditability. Under the hood, Grok 3’s reinforcement-learning–enhanced training enables multi-step reasoning across mathematics, coding, and analysis. As it emphasizes depth, Think typically uses more compute/latency than default replies. The underlying Grok-3 spec offers a 131,072-token context window, function calling, and structured outputs; pricing on the API rate card is $3/M input tokens and $15/M output tokens. End-user access is offered in X’s Grok experience (Premium+ / SuperGrok tiers), while developers can use Grok-3 via the xAI API.
Text
Overview
Grok 3 Think is xAI’s reasoning-tuned mode of Grok 3 that “thinks before responding” and can expose a step-by-step reasoning trace. It targets harder math, coding, and analysis, trading speed for depth. Built on the Grok-3 model (131,072-token context, function calling, structured outputs), it’s available via the xAI API and to X Premium+/SuperGrok users
About xAI
xAI is an artificial intelligence startup founded by Elon Musk, aiming to understand the universe.
Industry:
Artificial Intelligence
Company Size:
1200
Location:
Palo Alto, California, US
Website:
x.ai
Benchmark scores
How Grok 3 Think ranks on tracked AI benchmarks. Click any benchmark to see its full leaderboard.
Compare across all benchmarks →Tools using Grok 3 Think
-
Grok — v4.5Handles difficult long-running tasks more effectively by investigating problems, using tools, recovering from mistakes, and verifying results in realistic environments. Extends beyond software engineering into data science, finance, legal work, and other computer-based knowledge tasks with a broader training mix. Improves agent-style computer work through training on large volumes of codebase interactions and developer-agent behavior with software tools. Gains stronger robustness on hard problems through reinforcement learning environments designed specifically to remain challenging even for frontier systems. Offers a separate fast variant, indicating a higher-speed option alongside the main weight class for lower-latency workflows.
Last updated: April 6, 2026
