xAI Launches Grok 4.6 With 500K Token Context Window

xAI Ships Grok 4.6

xAI has released Grok 4.6, the latest version of its Grok model family, now accessible via OpenRouter's API under the identifier x-ai/grok-4.6.

The headline specification is a 500,000-token context window, a substantial jump that puts the model in the upper tier of context length among commercially available frontier models. Grok 4.6 accepts text, images, and files as input and produces text output, placing it in the multimodal category.

Pricing

Grok 4.6 is priced at $2.00 per million input tokens and $6.00 per million output tokens. That output price is three times the input price, a ratio consistent with pricing structures used by other frontier-model providers.

Claims on Capability

According to xAI, Grok 4.6 is the company's "smartest model" to date, with what it describes as frontier performance on coding, knowledge work, and STEM tasks. xAI has not published specific benchmark scores (e.g., MMLU, HumanEval, or GPQA results) alongside this release, so these performance claims remain unverified pending independent testing or official benchmark disclosures.

No training data cutoff date, parameter count, or architecture details have been disclosed for Grok 4.6.

Availability

The model is live now through OpenRouter, which functions as a routing layer giving developers unified API access to multiple model providers. This makes Grok 4.6 immediately usable by any application already integrated with OpenRouter's infrastructure, without requiring a direct integration with xAI's own API.

What This Means

A 500K token context window is large enough to process lengthy codebases, multi-document research collections, or extended conversation histories in a single call — a meaningful practical upgrade for developers building agents or retrieval-heavy applications, regardless of how the underlying model performs on standard benchmarks.

The $2/$6 per-million-token pricing positions Grok 4.6 competitively against other frontier multimodal models, though absent independent benchmark data, it's not yet possible to verify xAI's claims of "frontier performance" relative to models like GPT-4-class or Claude systems. Buyers evaluating Grok 4.6 for production use should request or run their own benchmark comparisons before committing, particularly on coding and STEM tasks where xAI's claims are most specific. The rapid availability through OpenRouter, rather than an exclusive xAI-only launch, suggests xAI is prioritizing developer reach over a walled-garden distribution strategy for this release.