Grok 4.7: a same-price flagship replacement, with the long-context surcharge unchanged

xAI released Grok 4.7 on September 21, 2026 as `grok-4.7`. It replaces 4.6 as the current API flagship, keeps the 500K context window and the same $2 input / $6 output per million tokens below 200K input, and adds stronger long-horizon coding, knowledge work and self-checking according to xAI. This is a version correction for our calculator, not a repricing.

Verdict: xAI ships a more reliable coding-and-knowledge flagship at the same $2/$6 rate; the calculator changes name, not price

The take

The rate card is the load-bearing source: $2.00 input, $0.50 cached input and $6.00 output per million tokens for requests below 200K input; $4 / $1 / $12 once a request crosses 200K, with the higher rate applying to the whole request. That is identical to Grok 4.6. We therefore preserve the calculator row key for old share links and rename the visible model to 4.7.

xAI describes 4.7 as its most capable model for coding and knowledge work, with better performance on longer tasks and additional self-checks. Its launch benchmark table is vendor-run; we have not independently reproduced those figures, so they do not move a local planner pick. The operational fact is firmer: 4.7 is live through the API and partner tools, while xAI's general model index was still lagging behind its own launch and pricing pages when we checked.

Our call: use 4.7 instead of 4.6 for new hosted work at the same budget, but keep the long-context cliff in view. Nothing local changed: this is closed, hosted software, and no consumer hardware recommendation is displaced.

Where this fits

Models: Qwen3.8-27B · DeepSeek V4-Flash · Kimi K3 (2.8T-A50B)

Hardware: NVIDIA RTX 5090 · Mac Studio M4 Max 64 GB

Sources

Next step

Try this in the planner→