Grok Code Fast 1 is Here!

Grok Code Fast 1 is a speedy reasoning model developed by xAI that excels at agentic coding.

From Sonic to Grok Code Fast 1

xAI recently released the Sonic model as a stealth model through select partners including Cursor, Windsurf, and other development platforms. This initial release allowed xAI to gather valuable feedback and refine the model before its another launch as Grok Code Fast 1. The model is currently available for free through various partners.

Key Stats

• Model Name: grok-code-fast-1
• Developer: xAI
• Release Date: August 26, 2025
• Context Window: 256k tokens
• Input Pricing: $0.20 per million tokens
• Output Pricing: $1.50 per million tokens
• Modalities: Text-to-text (image input not currently supported)
• Optimized for: Daily coding use, agentic workflows

What is Grok Code Fast 1?

Grok Code Fast 1 is a small coding optimized AI model designed for daily use in coding and development tasks. It offers rapid reasoning capabilities with visible thought processes, making it ideal for developers who need quick, high-quality code assistance.

Small Coding Optimized Model

As emphasized by xAI's CEO Elon Musk, Grok Code Fast 1 is specifically designed as a smaller, coding-optimized model that provides efficient performance for development tasks:

Grok Code, which comes out soon, is a smaller coding-optimized model

— Elon Musk (@elonmusk) August 19, 2025

Model Specifications

Context Window

Grok Code Fast 1 features a large 256k context window, allowing it to handle extensive codebases and maintain context across long conversations and complex coding sessions.

Performance

• Speed: Optimized for fast response times
• Reasoning: Advanced reasoning capabilities with visible traces
• Agentic Coding: Excels at autonomous coding tasks and workflows

Capabilities

• Function Calling: Connect the xAI model to external tools and systems
• Structured Outputs: Return responses in specific, organized formats
• Reasoning: The model thinks before responding with visible thought processes

Availability

Free Access Period

Grok Code Fast 1 is available for free until September 2, 2025, 1pm PT (within capacity limits) for some providers like Cursor. This makes it accessible for developers and teams looking to enhance their coding workflows without immediate cost concerns and also allows xAI to test and evaluate the model.

xAI API

Grok Code Fast 1 is now available for General Availability (GA) in xAI's direct API.

API Access Details

• Endpoint: Available through xAI's API endpoints
• Authentication: Standard xAI API authentication required
• Caching: Supports input token caching to reduce costs
• Function Calling: Full support for connecting to external tools and systems

Use Cases

Grok Code Fast 1 is particularly well-suited for Coding!

Where to Access

Grok Code Fast 1 is available across multiple platforms with various free access periods:

Kilo Code

We've teamed up with xAI to give you unlimited free access to Grok Code Fast, inside Kilo Code.

Kilo Code is a free, open-source AI coding extension for VS Code that hit 240,000 downloads in under 4 months.

Here's how to get unlimited free access to Grok Cost Fast: pic.twitter.com/fOdxOI3E7v

— Kilo Code (@Kilo_Code) August 26, 2025

Cursor

Grok Code is now available in Cursor! It's free for one week.

It's fast and has competitive pricing ($0.2/1M input, $1.5/1M output) pic.twitter.com/kksbSwxYFe

— Lee Robinson (@leerob) August 26, 2025

Windsurf

Grok Code Fast 1 is now available for free in Windsurf for a limited time. Let us know what you think! pic.twitter.com/JwqIWvxILZ

— Windsurf (@windsurf) August 26, 2025

opencode

Grok Code

Free on opencode

For a limited time pic.twitter.com/XTDGokRWwW

— SST (@SST_dev) August 26, 2025

OpenRouter

92 tokens per second (one of the fastest coding reasoning models on OpenRouter)

Stats are from OpenRouter.

xAI Provider Statistics

• Region: US
• Latency: 1.12s
• Throughput: 108.8tps
• Uptime: 100.0%

Model Specifications

• Total Context: 256K
• Max Output: 10K
• Input Price: $0.20
• Output Price: $1.50
• Cache Read: $0.02
• Cache Write: --
• Input Audio: --
• Input Audio Cache: --

Fifth model launch of the day‼️Grok Code Fast, from @xAI

At 92 tokens per second so far, it's one of the fastest coding models on OpenRouter, and comes with 256k context pic.twitter.com/e0k6nh5XyX

— OpenRouter (@OpenRouterAI) August 26, 2025

Additional Platforms

• Other supported platforms and APIs

Pricing

• Input tokens: $0.20 per million tokens
• Cached input tokens: $0.02 per million tokens
• Output tokens: $1.50 per million tokens
• Live search: Not supported

Rate Limits

• Requests per minute: 480
• Tokens per minute: 2,000,000

Model Details

• Model name: grok-code-fast-1
• Aliases: grok-code-fast, grok-code-fast-1-0825
• Region: us-east-1

Historical Context

Just a few days before the Grok 4 announcement, some information about the Grok 4 Code model was surfaced through various channels.

Grok 4 Code

• Display Name: Grok 4 Code
• Model ID: grok-4-code-0629
• Tagline: Engineering Intelligence Unleashed
• Description: A model purpose-built to be your coding companion. Ask it questions about your code or embed directly into your code editor

Previous Identity: Grok 3.5 Code

• Previous Model Name: Grok 3.5 Code
• Previous Model ID: grok-3-5-code-0621
• Previous Title: Engineering Intelligence Unleashed
• Previous Description: Our reasoning model that specifically targets software engineering tasks

Additional Historical Model References

• grok-3-5-code-0615
• grok-3-5-code-0625
• grok-4-code-0630
• grok-4-code-0702

Benchmark Results

These Results are Not Final and Not Official, however they were surfaced through various channels from xAI's resources. This results were surfaced on July 4th, 2025.

Note: Grok Code Fast 1 is released on August 26th, 2025 and it's not confirmed if it's the same model as Grok 4 Code.

Grok 4 Code (grok-4-code-0629)

Standard Benchmarks

• SWE-Bench: 72
• Terminal Bench: N/A

Test Time Compute Benchmarks

• SWE-Bench: 75
• Terminal Bench: N/A

Grok Code Fast 1 is Extremely Fast and Super Capable coding reasoning model. Try it out!