xAI releases Grok 4.7 for long-running coding work

xAI has released Grok 4.7 as a frontier model for coding, agentic tasks, and knowledge work. The model is available through the xAI API as grok-4.7, with text and image input, a 500,000-token context window, configurable reasoning effort, and no text output limit documented on the release notes.

The launch also changes the cost conversation for long-running coding workloads. xAI lists pricing from $2 per million input tokens and $6 per million output tokens below 200,000 prompt tokens, with higher rates above that threshold. Vercel lists Grok 4.7 in AI Gateway, fx, and eve, with a 40% discount advertised for one week from September 21.

Grok 4.7 targets longer agent runs

xAI says the model uses a larger base model than Grok 4.6 and a longer reinforcement-learning run weighted toward tasks that take hours rather than minutes. The release highlights code verification, longer context management, document and presentation creation, and performance on coding, terminal, legal, and clinical-reasoning evaluations.

Those claims describe xAI’s own benchmark framing. They are useful for understanding the intended workload, not a substitute for testing a model on a team’s repository, tool permissions, latency budget, and failure-handling requirements.

API access comes with concrete limits

The API model accepts text and image inputs and returns text. Reasoning effort can be set to low, medium, high, or xhigh. The documented price tiers change when a prompt exceeds 200,000 tokens, so cost estimates for large codebases should model both the input and output sides instead of using a single headline rate.

The release also says Grok 4.7 is available in Cursor, Grok Build, third-party coding harnesses, model routers, and cloud platforms. Vercel’s changelog independently confirms availability in AI Gateway and names a temporary discount window. Availability through a router does not guarantee identical latency, retention, regional routing, or tool behavior; those details depend on the provider selected behind the interface.

What builders should verify first

Teams evaluating Grok 4.7 should pin the exact model identifier, record the reasoning setting, and check the provider’s current price and data-handling terms. For agent workflows, test long-running tasks with explicit tool budgets and inspect whether the model’s self-checking improves recovery from partial failures or simply adds more tokens.

The Vercel discount is time-limited, while xAI’s API pricing and availability are documented separately. The next practical checkpoint is whether the model’s claimed price-performance advantage survives independent workload tests after the promotional window ends.

Sources and methodology

This report uses xAI’s Grok 4.7 announcement as the primary source and Vercel’s public changelog as independent corroboration for platform availability and the dated discount. The article preserves the release stage as available now and separates xAI’s benchmark claims from conclusions that require buyer-side testing.