Grok 4: Pricing, Benchmarks & Access Guide [2026]

xAI just rolled out Grok 4, its most powerful model yet, and it’s now available to both developers (via API) and everyday users.

This release features significant enhancements to context length, multimodal capabilities, speed, and API pricing.

If you're after performance, scale, and API access without the red tape, Grok 4 is a strong new option worth exploring.

What Is Grok 4?

Grok 4 is a multimodal large language model from xAI, designed to handle both image and text input with high reasoning accuracy and long-context processing.

Compared to Grok 3, this release is a clear upgrade:

• Context length: 256,000 tokens
• Inputs: Text and image
• Output: Text-only
• Reasoning: Always-on (can’t be disabled or toggled)

xAI has positioned Grok 4 as a reasoning-first model. It breaks down queries into sub-steps, processes information more deliberately, and produces structured answers. You can’t view or turn off its internal reasoning mode, but the results are generally more accurate—especially for technical tasks and abstract queries.

Independent testing shows Grok 4 scoring 73 on the Artificial Analysis Intelligence Index, outperforming:

• OpenAI o3 (score: 70)
• Gemini 2.5 Pro (70)
• Claude 4 Opus (64)
• DeepSeek R1 0528 (68)

It also handled image inputs well. In one real-world test, Grok 4 generated an SVG image of a pelican riding a bicycle, and then successfully described the same image using its image analysis capabilities.

But it’s not all smooth. Grok 4 launched without a model card, and recent issues with Grok 3’s prompt alignment (including an embarrassing MechaHitler incident) suggest xAI still has work to do on safety and trust.

Still, in terms of speed, reasoning, and raw capability, Grok 4 stands out.

Grok 4 is xAI's most capable model, but deploying it for real customer interactions requires custom training and multi-channel support. Over 10,000 businesses use Chatbase to bridge that gap in under 30 minutes.

Try Grok 4 through Chatbase for free

How to Access Grok 4

Accessing Grok 4 is flexible, with multiple options designed for both everyday users and developers:

1. SuperGrok Subscription on X(Consumer Access)

Platform: Grok 4 is available directly within X (formerly Twitter). Access: Available to users subscribed to SuperGrok plans. How to Access: Once subscribed, users can interact with Grok 4 directly on the X mobile app or desktop platform. Plans:

• SuperGrok: $30/month or $300/year – access to Grok 4
• SuperGrok Heavy: $300/month or $3,000/year – access to Grok 4 Heavy (a larger variant)

Grok 4 Pricing: How Much Does It Cost?

Grok 4 is available across several pricing tiers depending on how you want to use it:

• SuperGrok: $30/month or $300/year. Gives you full access to Grok 4 through the X platform and grok.com.
• SuperGrok Heavy: $300/month or $3,000/year. Access to Grok 4 Heavy, a larger variant with enhanced reasoning for the most demanding tasks.
• API (standard context, up to 128K tokens): $3 per million input tokens, $15 per million output tokens.
• API (extended context, over 128K tokens): $6 per million input tokens, $30 per million output tokens.

There is no free tier for Grok 4 itself. The free access on X applies to Grok 3 only. If you want to try Grok 4 without committing to a subscription, Chatbase offers access through its platform where you can test multiple models including Grok 4 on a free plan.

2. Standalone Web Interface

Website: Grok.com offers a dedicated web-based experience. Access: Access may still be limited in some regions (such as the EU or UK), but rollout is ongoing. Functionality: Full access to Grok 4’s capabilities, depending on subscription tier.

3. Mobile Applications

Platforms: Grok 4 is available on iOS and Android. Availability: App availability varies by country, with support expanding progressively. Access: Linked to your X subscription; SuperGrok subscribers can access Grok 4 through the official X app.

4. API Access

xAI has released API access to Grok 4, making it available for developers to integrate into applications and workflows.

Pricing (standard context window ≤ 128K tokens):

• $3 per million input tokens
• $15 per million output tokens

Pricing (extended context > 128K tokens):

• $6 per million input tokens
• $30 per million output tokens

The pricing model is simple and predictable, comparable to OpenAI's API pricing, and ideal for scaling projects without cost surprises.

You can start building with Grok 4 right now via Chatbase—no waitlist, no extra setup.

→ Access Grok 4 on Chatbase

With Chatbase, developers can train and fine-tune Grok 4 using custom data before deploying it into production.

Chatbase's Retrieval-Augmented Generation (RAG) infrastructure helps you build AI assistants that are more accurate, deeply personalized, and ready for real-world use

Whether you're automating customer support, building internal tools, or integrating AI into your product, Grok 4 is ready to go.

→ Want to train Grok 4 with your own data? Sign up on Chatbase and start building today.

A Timeline to Grok 4: From Grok-1 to Now

xAI hasn’t been around as long as OpenAI or Google DeepMind, but it’s moved quickly. In just over a year, it’s gone from launching its first model to shipping Grok 4, a system now competing head-to-head with the top players in the space.

Despite concerns around prompt safety and the lack of a formal model card at launch, Grok 4 represents xAI’s most complete, production-ready model to date.

• November 2023 – Grok-1 Launches xAI introduced its first model, Grok-1, integrated directly into X (formerly Twitter). It showed promise but lacked the power and refinement of leading models like GPT-4.
• March 2024 – Grok-1 Goes Open Source xAI open-sourced Grok-1 under the Apache-2.0 license. This gave developers the ability to experiment freely and marked a bold move toward transparency and speed.
• August 2024 – Grok-2 Arrives Grok-2 delivered better reasoning and introduced image generation for the first time. It signaled real progress in xAI’s development pipeline.
• October 2024 – Grok-2 Gains Vision Grok-2 was updated to understand images, not just generate them. This made it more competitive in multimodal AI tasks like content analysis and education.
• December 2024 – Platform Expansion Begins Grok started rolling out standalone access via web and iOS apps, beginning with users in Australia. It was no longer just a chatbot inside Twitter—it was becoming a platform.
• February 2025 – Grok 3 Launches Grok 3 was a big leap forward, powered by xAI’s Colossus supercomputer. It introduced advanced reasoning modes like “Think” and “Big Brain,” and user numbers surged.
• July 2025 – Grok 4 Goes Live Grok 4 launched with 256K context length, text+image input, and strong benchmark performance. It’s now available via API, mobile apps, and the SuperGrok subscription plan, marking xAI’s most production-ready model to date.

What Is Grok 4.20 (Grok 4 Heavy)?

Grok 4.20, also referred to as Grok 4 Heavy, is the larger variant of Grok 4 designed for tasks that require deeper reasoning and more compute. It is available through the SuperGrok Heavy subscription at $300/month.

In benchmark testing, Grok 4.20 Heavy has shown improvements over the standard Grok 4 model on complex multi-step reasoning tasks, though specific benchmark numbers from xAI have not been published separately from the base model. Independent testers on LM Arena have noted that the Heavy variant handles longer chains of reasoning more reliably, particularly on math and coding challenges.

For most users, the standard Grok 4 model is sufficient. Grok 4 Heavy makes sense for professional use cases where you need the absolute highest accuracy on complex technical problems and are willing to pay 10x the standard subscription cost.

Why Choose Grok 4?

Grok 4 stands out in a crowded field of LLMs by combining strong performance with developer flexibility and competitive pricing. Here’s what makes it worth considering:

• 256K context length, double that of Grok 3, ideal for handling large documents or deep multi-step prompts
• Multimodal input, allowing you to feed in both text and images for more versatile tasks
• Always-on reasoning, delivering structured, thoughtful responses without needing to toggle special modes
• Live data awareness, with the ability to reference real-time content from X (formerly Twitter), including actual tweets
• Straightforward pricing, starting at $3/million input tokens and $15/million output tokens—with extended context pricing clearly defined
• Flexible access options, from API (OpenRouter, Chatbase) to mobile apps and SuperGrok subscriptions on the X platform
• Production-ready performance, validated by strong benchmark scores, even as xAI continues to address model alignment and safety.

You have seen the benchmarks and pricing. Chatbase lets you deploy Grok 4 with RAG, custom training, and multi-channel support, all without managing API infrastructure yourself.

Build my AI agent with Grok 4

Grok 4 vs. Grok 3: What Changed?

If you are currently using Grok 3, here is what Grok 4 changes:

• Context window: Grok 4 has a 256K token context window. Grok 3 has 1 million tokens. Grok 3 still wins on raw context length.
• Multimodal input: Grok 4 accepts both text and image input. Grok 3 is text-only.
• Reasoning: Grok 4 has always-on reasoning that cannot be toggled off. Grok 3 has optional extended thinking mode. Grok 4's approach produces more consistently structured answers.
• Benchmarks: Grok 4 scores 73 on the Artificial Analysis Intelligence Index, outperforming GPT o3 (70) and Gemini 2.5 Pro (70). Grok 3 scored lower on the same index.
• Pricing: Both share the same API pricing ($3/$15 per million tokens). SuperGrok subscription covers both.

Bottom line: Grok 4 is better for tasks that benefit from image understanding and deeper reasoning. Grok 3 is better if you need the largest possible context window (1M tokens) for processing very long documents.

Ready to Get Started with Grok?

Grok 4 might not be the loudest model on the market, but it’s shaping up to be one of the most capable—especially if you care about reasoning quality, input flexibility, and predictable pricing.

It brings real improvements over Grok 3, performs well in independent benchmarks, and is already being used across a growing number of developer platforms.

Whether you’re building AI tools, automating customer support, or experimenting with next-gen assistants, Grok 4 is worth putting into your workflow.

→ Want to start using Grok 4 now? You can try it instantly on Chatbase, where you can train it on your own data and deploy it without setup delays.