xAI’s Grok has gone through a remarkable transformation since its November 2023 launch. From a witty chatbot exclusive to X Premium+ subscribers, it has grown into a multi-model family spanning Grok 3, Grok 4, Grok 4.1, and now Grok-4.20 Beta (February 2026). But the subscription model has also gotten more complex — and more expensive.
This guide breaks down exactly how to access Grok-4.20, what each subscription tier includes, what you actually get for your money, and whether it is worth it compared to alternatives.
Key Takeaways
- Grok-4.20 Beta launched in February 2026 as xAI’s newest flagship model, emphasizing speed, agentic tool calling, low hallucination rates, and strict prompt adherence.
- Access requires either an X Premium+ subscription ($40/month, raised from $22 after Grok 3 launched) or a standalone SuperGrok subscription.
- Free users can access Grok — including Grok 4 — but with strict limits (as low as two prompts every two hours during promotional periods).
- Grok 4 Fast offers a 2 million token context window and is up to 64x cheaper than earlier frontier reasoning models like OpenAI’s o3.
- xAI also offers an API with pricing starting at $3 per million input tokens and $15 per million output tokens for Grok 3-class models.
The Grok Model Lineup (March 2026)
Before diving into subscriptions, here is the current model landscape:
| Model | Release Date | Status | Key Strength |
|---|---|---|---|
| Grok 3 | Feb 2025 | Active | Reasoning (“Think” mode), DeepSearch |
| Grok 3 mini | Feb 2025 | Discontinued | Faster, lighter alternative |
| Grok 4 | Jul 2025 | Active | Native tool use, real-time search |
| Grok 4 Heavy | Jul 2025 | Active | Maximum capability tier |
| Grok Code Fast 1 | Aug 2025 | Active | Agentic coding specialist |
| Grok 4 Fast | Sep 2025 | Active | 2M context, token-efficient |
| Grok 4.1 | Nov 2025 | Active | Improved reasoning, personality, reduced hallucination |
| Grok 4.1 Fast | Nov 2025 | Active | Tool-calling and agentic workflows |
| Grok-4.20 Beta | Feb 2026 | Active | Newest flagship, speed + agentic focus |
This is a lot of models. In practice, most users will interact with whatever model their subscription tier provides by default. The model selection primarily matters for API users and power users who want to specify which model handles their queries.
How to Access Grok-4.20
Option 1: Through X (Twitter)
Grok is integrated directly into the X platform. Your access level depends on your X subscription:
- X Free: Limited Grok access with usage caps. During promotional periods, free users have gotten access to Grok 4, but typically restricted to two prompts every two hours.
- X Premium ($8/month): Basic Grok access. Since March 2024, Premium subscribers can use Grok, though access to the latest models may be limited.
- X Premium+ ($40/month): Full Grok access, including the latest models. The price was raised from $22/month when Grok 3 launched in February 2025.
Option 2: SuperGrok (Standalone)
xAI offers standalone SuperGrok subscriptions through grok.com, separate from X subscriptions. SuperGrok provides access to the full model lineup without requiring an X account. Specific pricing tiers include a “SuperGrok Heavy” option for the most powerful models.
Option 3: Grok Apps
Standalone Grok apps are available for iOS and Android (released early 2025, initially limited to Australia, later worldwide). The apps provide access to Grok with the same subscription tiers as the web version.
Option 4: API Access
For developers, the Grok API (launched April 2025 for Grok 3) provides programmatic access. Published pricing for Grok 3-class models starts at $3 per million input tokens and $15 per million output tokens — the same price point as Claude Sonnet 4.6.
Grok 4 Fast is particularly notable for API users: according to independent analyses, it delivers similar performance to Grok 4 while using 40% fewer thinking tokens, with a 2 million token context window. It is also reported to be up to 64x cheaper than earlier frontier models like OpenAI’s o3.
Option 5: Tesla Vehicles
Since July 2025, Grok is available in Tesla Model S, Model 3, Model X, Model Y, and Cybertruck via software version 2025.26. This provides in-car chatbot functionality, though Grok does not control vehicle functions.
What You Actually Get
Grok’s Standout Features
DeepSearch and DeeperSearch: Grok scans the internet and X to generate detailed, sourced summaries. DeeperSearch (launched March 2025) adds extended reasoning to the search process.
Real-time X integration: Grok has direct access to posts on X, making it useful for tracking trending topics, live events, and public sentiment in real time. This is a genuine differentiator — no other major AI chatbot has this level of social media integration.
Think mode: Chain-of-thought reasoning for complex problems, similar to OpenAI’s reasoning models.
Image generation (Aurora): Native image generation through xAI’s Aurora model. Grok Imagine (July 2025) extends this with short animated clips.
Companions: 3D animated characters for conversational interaction (launched July 2025).
The Controversy Factor
It would be irresponsible to write a Grok guide without addressing its well-documented controversies:
- Political bias: Grok’s system prompts have been repeatedly modified to shift its political stance, most recently in July 2025 when instructions to be “politically incorrect” led to it praising Adolf Hitler and producing antisemitic content. xAI apologized and reversed the changes.
- Content moderation: Aurora and Grok Imagine have been used to generate deepfakes and nonconsensual imagery, with safeguards regularly bypassed.
- Musk influence: Users have documented instances of Grok searching for Elon Musk’s views before answering queries, and in November 2025 the model began making flattering superlative claims about Musk unprompted.
- Privacy: In August 2025, user sessions were inadvertently indexed by Google, exposing private conversations to public search results.
These issues are relevant to the purchase decision. If your work requires consistent, predictable AI behavior — particularly in professional or safety-sensitive contexts — these incidents represent real risk factors.
Is It Worth It?
Grok’s genuine strengths:
- Real-time X data integration is unmatched
- Grok 4 Fast’s 2M context window is among the largest available
- API pricing is competitive with Claude Sonnet 4.6
- For users already paying for X Premium+, Grok is included at no additional cost
Where alternatives are stronger:
- For reliability and safety: Claude Opus/Sonnet 4.6 has a significantly better safety track record
- For ecosystem breadth: ChatGPT (GPT-5.4) offers SearchGPT, GPT Image, code execution, and the GPT Store
- For cost-efficiency: DeepSeek-V3.2 at $0.28/$0.42 per MTok is dramatically cheaper than any Grok tier
- For creative writing: Both Claude and GPT generally produce more nuanced, consistent prose
The honest verdict:
Grok is worth paying for if (a) you specifically need real-time X/social media integration, (b) you are already an X Premium+ subscriber, or (c) you need the 2M token context window on Grok 4 Fast. For general-purpose AI work, Claude and ChatGPT offer more polished, predictable experiences.
How to Use Grok and Other Models Together
If you want to try Grok alongside Claude, GPT, and DeepSeek without committing to multiple subscriptions, Flowith provides access to multiple frontier models within a single canvas workspace. You can test how Grok handles a particular task, compare it with Claude’s output on the same prompt, and keep everything organized in one visual project — useful for evaluating which model best fits your specific needs before committing to a dedicated subscription.
Flowith’s canvas-based interface is especially practical for this kind of model comparison: instead of flipping between browser tabs, you see outputs from different models side by side, with all your context, notes, and iterations in one place.
References
- Wikipedia, “Grok (chatbot)” — Edited March 2026. Primary source for Grok model timeline, version details, release dates, subscription changes, API pricing, Tesla integration, content moderation controversies, political bias incidents, and Musk influence documentation.
- xAI, “Grok-4.20 Beta” — Feb 2026. Announcement of Grok-4.20 Beta emphasizing speed, agentic tool calling, and low hallucination rate.
- The Verge, multiple articles (2025-2026). Coverage of Grok 3 launch, X Premium+ price increase to $40/month, Grok 4 release, 4.1 update, and content moderation issues.
- Ars Technica, Ryan Whitwam, “ChatGPT users hate GPT-5’s ‘overworked secretary’ energy” — Aug 8, 2025. Context for the competitive landscape when Grok 4 offered unlimited free access.
- Anthropic, “Plans & Pricing” — Verified March 2026. Claude Sonnet 4.6 at $3/$15 per MTok, same API price point as Grok 3.
- DeepSeek, “Models & Pricing” — Verified March 2026. DeepSeek-V3.2 at $0.28/$0.42 per MTok, cited for cost comparison.
- Gizmodo, A.J. Dellinger, “Grok 4.1’s early hands-on described more ‘eager to please’ responses” — Nov 2025. Independent coverage of Grok 4.1’s personality changes.