xAI's heavy-reasoning model. Real-time X integration, distinctive voice, premium price.
Grok 4 was xAI's first serious push into the frontier reasoning tier. It's still on the price list at $3 input and $15 output per 1M tokens with a 256K window — the same shape as Claude Sonnet 4.6 and GPT-5.4 — but the spotlight has moved to Grok 4.20, the 2026 flagship that ships at $2 input and $6 output. Grok 4 remains worth choosing when you want xAI's specific reasoning style at premium output quality.
What you really pay for with the Grok family is real-time access to the X (Twitter) timeline and a model voice that's noticeably less filtered than the OpenAI and Anthropic equivalents. For some workloads that's a feature; for compliance-bound enterprises, it's a liability.
Grok 4 is a strong reasoner, particularly on math and structured logic, and a competent generalist on writing and code. The Deep Search tool surfaces live web and X content during a request, which makes Grok the easiest way to ask "what happened on X in the last hour" without writing your own scraping layer. Vision and image editing are supported; image generation is delegated to other xAI products.
Weaknesses: tool-use ergonomics lag the OpenAI Responses API, and the model's safety tuning is opinionated in ways your end users may or may not appreciate. Pick deliberately.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Grok 4 | $3 | $15 | 256K |
| Grok 4.20 | $2 | $6 | 256K |
| Grok 4.1 Fast | $0.20 | $0.50 | 2M |
| Claude Sonnet 4.6 | $3 | $15 | 200K (1M β) |
If you don't specifically need Grok 4, the newer Grok 4.20 is cheaper and faster. Versus Sonnet 4.6 — the closest cross-family rival on price — Grok 4 trades writing polish for real-time X data and a different model voice.