Breaking

xAI Has Not Shipped Grok 4.7 After Musk Cites Extra Days of RL Tuning

September 13, 2026 at 16:05 EDT

Foundation Models
Software Dev & Coding

xAI has not released Grok 4.7, its next flagship language model, after Elon Musk said on September 11 that the system still needs a few more days of reinforcement-learning work. As of September 13, the company has published no launch note, API identifier, price, or benchmark card for the successor, and Grok 4.6 remains the documented production model.

September 13, 2026 · xAI

Grok 4.7 is close. It is not shipped.

Elon Musk said the next Grok flagship needs a few more days of reinforcement-learning tuning after public windows in July, August, and early September closed. xAI has published no model page, price, or scores.

Claimed parameter count

Musk has described Grok 4.7 as a 2.1 trillion parameter model, versus 1.5 trillion for Grok 4.6 — a 40% jump that xAI has not put on a spec sheet.

1.5T

Grok 4.6

2.1T

Grok 4.7 (claimed)

Vertical stacks scaled 5 to 7, matching 1.5T to 2.1T.

+40%

Stated size vs Grok 4.6

$2 / $6

Grok 4.6 API, per 1M tokens

Aug 12

Grok 4.6 still the live flagship

Sept 11

"A few more days" delay note

Path to a still-unreleased model

Initial training done

Musk, Aug 12

→

SpaceX data added

Supplemental training

→

RL length tuning

Quits hard tasks early

→

Not yet shipped

No ID, price, or scores

Why some want the wait

Builders watching the launch say it is better to hold the model than ship one that still abandons hard problems or skips self-checks.

Why others are skeptical

Public targets from July, August, and early September all closed. Some argue a few days will not fix a foundation-model behavior gap.

Still missing from xAI

Grok 4.7 is a founder description, not a callable product.

No grok-4.7 model IDNo published priceNo context windowNo official scores

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.