# Grok 4 Fast

> xAI's cost-efficient Grok 4 with a 2M-token context window.

- Maker: xAI (Grok)
- Line: Grok Fast
- Canonical page: https://ai-tldr.dev/models/grok-4-fast/
- Updated: 2026-06-18

## In short

Grok 4 Fast is xAI's cost-optimized member of the Grok 4 family, released on September 19, 2025. It has a 2M-token context window and costs $0.20 per million input tokens and $0.50 per million output tokens. You can use it through xAI and OpenRouter.

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 2M | [1] |
| Input price | $0.20 / 1M tokens (cached $0.05 / 1M tokens) | [2] |
| Output price | $0.50 / 1M tokens | [2] |
| Licence | Proprietary · API only | [1] |
| Released | September 2025 | [3] |
| Parameters | Undisclosed | [1] |
| Knowledge cutoff | Not publicly disclosed | [1] |

## Sources

1. [Grok 4 Fast API pricing & providers — PricePerToken](https://pricepertoken.com/pricing-page/model/xai-grok-4-fast)
2. [Grok 4 Fast (Reasoning) — Artificial Analysis](https://artificialanalysis.ai/models/grok-4-fast-reasoning)
3. [xAI Models documentation](https://docs.x.ai/developers/models)

## Overview

Grok 4 Fast is xAI's cost-optimized member of the Grok 4 family, released on September 19, 2025. It is designed to deliver close-to-frontier quality at a fraction of the price of the full Grok 4 model, pairing aggressive token efficiency with a very large 2 million-token context window. xAI markets it as its most cost-efficient model, aimed at high-volume production workloads, search, and agent loops where price per answer matters.

What sets Grok 4 Fast apart architecturally is its unified design: a single set of weights serves both a reasoning mode (extended chain-of-thought) and a fast non-reasoning mode, exposed through the grok-4-fast-reasoning and grok-4-fast-non-reasoning endpoints. xAI reports that Grok 4 Fast reaches Grok 4-level scores on several benchmarks while spending roughly 40% fewer thinking tokens, which combined with low per-token rates makes it dramatically cheaper to run.

Grok 4 Fast accepts text and image inputs and returns text, with native tool use including web and X (Twitter) search and link-following for grounded, up-to-date answers. API pricing starts at $0.20 per million input tokens and $0.50 per million output tokens, with cached input at $0.05 per million, placing it among the cheapest frontier-adjacent APIs available at launch.

## Benchmarks

- GPQA Diamond: 85.7% ([source](https://binaryverseai.com/grok-4-fast-review-benchmarks-pricing-2m-context/))
- AIME 2025 (no tools): 92% ([source](https://binaryverseai.com/grok-4-fast-review-benchmarks-pricing-2m-context/))
- HMMT 2025 (no tools): 93.3% ([source](https://binaryverseai.com/grok-4-fast-review-benchmarks-pricing-2m-context/))
- LiveCodeBench (Jan-May): 80% ([source](https://binaryverseai.com/grok-4-fast-review-benchmarks-pricing-2m-context/))
- Humanity's Last Exam (no tools): 20% ([source](https://binaryverseai.com/grok-4-fast-review-benchmarks-pricing-2m-context/))

## Pricing

- Input: $0.20 / 1M tokens per 1M tokens
- Cached input: $0.05 / 1M tokens per 1M tokens
- Output: $0.50 / 1M tokens per 1M tokens

Base rates for context up to 128K tokens; a higher tier applies above 128K. Verify current rates in the xAI console.

Source: https://artificialanalysis.ai/models/grok-4-fast-reasoning

## Strengths

- Very large 2 million-token context window for long documents, codebases, and multi-step agent histories
- Low API cost ($0.20 input / $0.50 output per million tokens, $0.05 cached) with strong benchmark scores
- Unified single-weights design exposes both fast and deep-reasoning modes without switching models
- High token efficiency — roughly 40% fewer thinking tokens than Grok 4 for comparable quality
- Native tool use with built-in web and X search for grounded, current answers
- Multimodal text + image input

## Best for

- High-volume production tasks where cost per answer is the deciding factor
- Long-context work: analyzing large documents, transcripts, or entire codebases within the 2M window
- Agentic search and research loops that benefit from native web and X search
- Real-time coding assistance and competitive-math-style reasoning
- Latency- and budget-sensitive chat and summarization at scale
- Multimodal tasks combining text and image inputs

## How to access

- [xAI](https://docs.x.ai/developers/models) — model id `grok-4-fast-reasoning`
- [xAI](https://docs.x.ai/developers/models) — model id `grok-4-fast-non-reasoning`
- [OpenRouter](https://openrouter.ai/x-ai/grok-4-fast) — model id `x-ai/grok-4-fast`

## Grok Fast versions

- [Grok 4.1 Fast](https://ai-tldr.dev/models/grok-4-1-fast/) (19 November 2025)
- Grok 4 Fast (September 2025) — this page

## FAQ

### When was Grok 4 Fast released?

xAI released Grok 4 Fast on September 19, 2025, as the cost-optimized member of the Grok 4 family.

### How big is Grok 4 Fast's context window?

Grok 4 Fast supports a context window of up to 2 million tokens, large enough for long documents, full codebases, and extended agent histories.

### How much does the Grok 4 Fast API cost?

Base pricing is $0.20 per million input tokens and $0.50 per million output tokens, with cached input at $0.05 per million. A higher tier applies for context above 128K tokens; check the xAI console for current rates.

### What is the difference between the reasoning and non-reasoning modes?

Grok 4 Fast uses a single set of weights that serves both modes. The reasoning endpoint (grok-4-fast-reasoning) does extended chain-of-thought for harder problems, while the non-reasoning endpoint (grok-4-fast-non-reasoning) answers quickly for simpler tasks.

## Links

- [Grok 4 Fast (Reasoning) — Artificial Analysis](https://artificialanalysis.ai/models/grok-4-fast-reasoning)
- [Grok 4 Fast benchmarks & pricing review — BinaryVerse AI](https://binaryverseai.com/grok-4-fast-review-benchmarks-pricing-2m-context/)
- [Grok 4 Fast API pricing & providers — PricePerToken](https://pricepertoken.com/pricing-page/model/xai-grok-4-fast)
- [Grok 4 Fast on OpenRouter](https://openrouter.ai/x-ai/grok-4-fast)
- [xAI Models documentation](https://docs.x.ai/developers/models)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/grok-4-fast/