Grok 4: Features, login, costs, alternatives

Grok 4 combines chat, image analysis, and file upload into a single tool that you can launch with a single click on your X account and use at fair rates.

Grok 4: Features and getting started

On July 10, 2025, xAI launched Grok 4. Compared to Grok 3, the release adds integrated image analysis, a 256,000-token context, the optional multi-agent mode “Grok 4 Heavy,” and now hidden internal reasoning; access and pricing remain the same.

• Open grok.x.ai, click on “Sign in with X” and confirm the access request; the single sign-on (SSO) routine will take your public profile name and immediately create your Grok profile.

• After logging in, you can access the chat window in the X sidebar or directly in your browser. There, you can paste text, image URLs, or JSON files into the same dialog and receive responses and files in one place.

• Grok 4 accepts up to 256,000 tokens per request, allowing you to process entire contracts, book chapters, or large code bases without splitting them up.

• Activate “Parallel Tools” in the side menu to perform multiple web searches, table analyses, or translations simultaneously; internal agents divide the tasks and consolidate the results.

• Drag and drop images or PDFs into the chat window; the model recognizes objects, reads diagrams, and describes content as long as your daily usage limit has not been reached.

• For data- or computation-intensive projects, select “Grok 4 Heavy” from the drop-down menu. This option launches multiple model instances, compares their responses, and provides a consolidated solution.

Costs, pricing, and alternatives

This section summarizes the current prices and shows you what alternatives you can consider.

• The standard subscription costs $30 per month and includes unlimited chats, image evaluation, and a personal API key; X Premium+ users get Grok 4 at no extra charge.

• SuperGrok Heavy costs $300 per month; the multi-agent version launches several specialized Grok instances in parallel, lets them pursue different solutions independently, and then compares the results as in a learning group. This ensemble approach increases accuracy and covers blind spots of individual agents, but requires more computing time, and each agent session counts separately toward your daily quota.

• For developers, xAI charges US$3 per 1 million input tokens and US$15 per 1 million output tokens; reused tokens can be reduced to US$0.75 through caching.

• As a direct alternative, Anthropic Claude Sonnet 4 offers a context length of 200,000 tokens at identical rates of $3/$15 per million tokens.

• OpenAI o4-mini-high charges $1.10 per million input tokens and $4.40 for output tokens with a context buffer of 128,000 tokens.

• Microsoft 365 Copilot integrates with Word, Outlook, and Teams for $30 per user per month (annual contract); bundled with Microsoft 365 Business Basic, pricing starts at $36.

• Perplexity Pro provides source-based answers for $20 per month, while Perplexity Max with unlimited Labs features is priced at $200.

• For local testing, download free models such as Llama 3 Instruct via Hugging Face; however, factor in your own server or cloud costs, as free GPU time is limited.