xai / grok-4-5
grok-4-5
Grok 4.5 is xAI’s flagship reasoning model for coding and long-context work. It supports tool calling and image understanding, returning detailed text answers.
Overview
Grok 4.5 is xAI’s flagship text reasoning model for coding, research, and knowledge work. It “thinks” before it answers, so it can handle multi-step tasks and longer prompts with fewer missed steps. You can attach images for visual Q&A, then get a single clear text response that explains what it found and what to do next. It also supports agent-style workflows, where it can browse the web when you enable live search, so answers can reflect recent information instead of only training-time knowledge. (docs.x.ai)
What you can build
- Bug fixing and code review that explains the root cause and a safe patch
- Long-context architecture reviews from large specs, logs, or chat history
- Visual Q&A on screenshots, UI mockups, charts, and photos
- Web-grounded research summaries when you need up-to-date facts
- Agentic workflows that mix reasoning with tools like web search and code execution (where supported by your setup)
Inputs
- A required written prompt that describes the task and any constraints.
- Optional image attachments for visual understanding. Use JPG/JPEG or PNG. Keep each image at or below 20 MiB.
- Optional system-level instructions that set behavior, tone, and boundaries for the whole session.
- Optional user identifier and session identifier to keep multi-turn chat history grouped under the right user and conversation.
- Optional reasoning depth selection (low, medium, or high). Higher depth spends more effort on harder problems.
- Optional live web search toggle for questions that need current facts.
- Optional output-length cap, so the response stops before it gets too long.
Outputs
Grok 4.5 returns a text answer. The response typically includes explanations, steps, and code snippets when you ask for them.
If you enable live web search, the response can also include source citations from pages it visited during the search process. (docs.x.ai)
Recommended settings
- Use low reasoning depth for quick tool use, lightweight coding help, and shorter chats.
- Use medium depth for data analysis, long-context reasoning, and careful writing.
- Use high depth for hard debugging, complex math, and multi-part planning.
- Turn on live web search when the question depends on recent events, changing facts, or fresh documentation.
Limitations
- Grok 4.5 always uses reasoning. You can reduce depth, but you can’t fully disable reasoning.
- Without live search tools, it won’t reliably answer questions about current events or very recent changes.
- Image understanding only supports JPG/JPEG and PNG. Oversized images (over 20 MiB) won’t work.
- Low-quality inputs can break results. Blurry screenshots, tiny text, heavy compression, or cluttered charts can hide key details.
- Availability can vary by region and account. As of July 8, 2026, xAI said Grok 4.5 was not yet available to EU users in the API console, with availability expected in mid-July 2026.
Safety & compliance
Follow xAI’s Acceptable Use Policy when you submit prompts, images, and generated output. Don’t try to bypass safety checks or generate disallowed content. Safety and moderation checks can still run even in stricter data-handling modes.
For data handling, xAI states it does not train on API inputs or outputs without explicit permission. xAI also describes a 30-day temporary storage window for API requests and responses for abuse auditing, with an enterprise option for zero data retention.
(x.ai)
API quick start
Run grok-4-5 with a single API call.
{
"prompt": "Tell me about Türkiye",
"inputImage": "https://your-cdn.com/input.png",
"user_id": "...",
"session_id": "..."
}