xAI Grok 4.1 Fast and Agent Tools API Release

xAI has introduced Grok 4.1 Fast and the Agent Tools API, providing developers with a specialized model and a server-side toolset designed for production-grade autonomous agents. This release focuses on improving tool-calling accuracy, reducing hallucinations, and integrating real-time data retrieval from the X platform.

Grok 4.1 Fast: Specialized Tool-Calling Model

Grok 4.1 Fast is engineered for enterprise use cases, specifically targeting complex agentic tasks in domains like finance and customer support. It features a 2-million-token context window and is optimized for speed and cost-effectiveness.

Performance Benchmarks

• $\tau^2$-bench Telecom: The model achieved a 100% score on this benchmark, which evaluates agentic tool use in customer support scenarios.
• Berkeley Function Calling v4: Grok 4.1 Fast achieved 72% overall accuracy with a total cost of $400.
• Factuality: The model reduces the hallucination rate by half compared to Grok 4 Fast, while maintaining performance on par with Grok 4 on the FActScore benchmark.

Long-Context Stability

To prevent the performance degradation typically seen as context length increases, xAI utilized long-horizon reinforcement learning with an emphasis on multi-turn scenarios. This ensures consistent performance across the full 2-million-token context window.

Agent Tools API: Server-Side Infrastructure

The Agent Tools API provides a suite of server-side tools that allow Grok 4.1 Fast to operate as an autonomous agent without requiring developers to manage their own API keys, sandboxes, or retrieval pipelines. The tools run entirely on xAI infrastructure.

Available Toolsets

• Search Tools: Provides real-time access to X posts and general internet search for current events and trends.
• Files Search: Enables intelligent retrieval of relevant documents from uploaded files, including citations.
• Code Execution: A secure sandbox for executing Python code to perform data analysis and and simulations.
• MCP Tools: Supports connection to external Model Context Protocol (MCP) servers for third-party tool integration.

Deep Research and Agentic Search

Combining Grok 4.1 Fast with the Agent Tools API enables state-of-the-art performance in deep research tasks. The model's native integration with the X ecosystem provides a significant advantage in real-time information retrieval.

Agentic Search Benchmarks

Benchmark	Grok 4.1 Fast Agent Tools API Score	GPT-5 Score	Claude Sonnet 4.5 Score	Gemini 3 Pro Score
Research-Eval Reka	63.9	45.5	41.2	55.9
FRAMES	87.6	86.0	85.0	90.9
X Browse*	56.3	24.2	14.6	26.5

*X Browse is an internal xAI benchmark evaluating multihop search and browsing capabilities on X.

API Availability and Pricing

xAI offers two variants of Grok 4.1 Fast via the API:

• grok-4-1-fast-reasoning: Optimized for maximal intelligence.
• grok-4-1-fast-non-reasoning: Optimized for instant responses.

Pricing Structure

Component	Price
Input Tokens	$0.20 / 1M tokens
Cached Input Tokens	$0.05 / 1M tokens
Output Tokens	$0.50 / 1M tokens
Tool Calls	From $5 / 1,000 successful invocations