No free plan. The cheapest way in is Input Tokens at $3.00 per million tokens. Consider free alternatives in the language model category if budget is tight.
The 2M token context is substantially larger than the context windows offered by most competing flagship reasoning models, which typically range from 128K to 200K tokens. This allows you to feed entire codebases, multi-volume documents, or extended conversation histories without chunking or retrieval-augmented workarounds. For long-context tasks like legal document review or full-repo refactoring, this is a meaningful advantage. However, retrieval quality at the upper end of any large context window varies, so empirical testing on your specific use case is recommended before committing.
Pricing is per-million-tokens: approximately $3.00/M for input tokens, $15.00/M for output tokens, $0.75/M for cached input tokens, and $5.25/M for image input tokens. The Artificial Analysis 'Price' metric blends input and output at a 3:1 ratio for fair cross-model comparison. There is no free consumer tier listed for direct API access; usage is metered and billed against an xAI account. For the latest rates, check xAI's API pricing page at x.ai or the live pricing comparison on Artificial Analysis, as per-token pricing updates periodically.
Artificial Analysis tracks it on the Intelligence Index v4.0, which aggregates 10 evaluations: GDPval-AA, Ī²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, and CritPt. These cover scientific reasoning, code execution, long-context retrieval, instruction following, and graduate-level domain knowledge. The composite index is designed to resist gaming by any single benchmark and provides a holistic view of model capability. Individual benchmark scores are also published for fine-grained comparison.
Yes â it supports both text and image inputs natively, making it a multimodal reasoning model rather than text-only. This enables use cases like chart interpretation, screenshot debugging, document OCR with reasoning, and visual question answering in a single API call. Image input is priced at approximately $5.25 per million tokens, separate from text token rates. Output is text-only; the model does not generate images.
Artificial Analysis measures output speed as tokens-per-second sustained after the first streaming chunk arrives, and tracks both median speed and variance over time. Grok 4.20 0309 v2 is highlighted for fast inference among comparable reasoning models, though absolute numbers vary by provider and load. Reasoning models typically have higher time-to-first-token than non-reasoning peers because they generate internal chain-of-thought before user-visible output. Check the Output Speed and Output Speed Over Time charts on Artificial Analysis for current measurements.
See Grok 4.20 0309 v2 plans and find the right tier for your needs.
See Pricing Plans âStill not sure? Read our full verdict â
Last verified March 2026