Alex Getmanearly AI signals
EN

Grok 4.7

xAI

Released Updated
Context window
500K
500,000 tokens
Input / Output
image, text → text

API pricing

USD / 1M tokens

Input lengthInputOutputCached input
Input ≤ 199,999 tokens$2.00$6.00$0.50
Input > 199,999 tokens$4.00$12.00$1.00

Thresholds include all input tokens, including cache hits.

Calculate your workload
Evidence links
Input
  • $2.00Input ≤ 199,999 tokens
  • $4.00Input > 199,999 tokens
Output
  • $6.00Input ≤ 199,999 tokens
  • $12.00Input > 199,999 tokens
Cached input
  • $0.50Input ≤ 199,999 tokens
  • $1.00Input > 199,999 tokens

Timeline

  1. · Product update · evidence link
Input / output price · USD / 1M tokens
ModelEvidence linksInput / output price
grok-4.7
standard · text
opencode-go$2.00 / $6.00
grok-4.7
batch
xai
API route unavailable
$1.60 (Input ≤ 199,999 tokens); $3.20 (Input > 199,999 tokens) / $4.80 (Input ≤ 199,999 tokens); $9.60 (Input > 199,999 tokens)
grok-4.7
standard
xai$2.00 (Input ≤ 199,999 tokens); $4.00 (Input > 199,999 tokens) / $6.00 (Input ≤ 199,999 tokens); $12.00 (Input > 199,999 tokens)
  1. · Evidence links · xAI · official articles
    Excerpt
    Grok 4.7 is available today in Cursor and Grok Build
  2. · Catalogue listing · GitHub · xai-org/grok-build · code
    Excerpt
    //! Immediate placement made grok-4.7 reason more in step replay; one call later it reasoned //! about a third less, so `delay` defaults to 1. use std::collections::VecDeque; use crate::util::config::LongReasoningReminderSettings; use xai_grok_config_types::BoolFlag; pub(crate) const ENV: &str = "GROK_LONG_REASONING_REMINDER"; pub(crate) const DEFAULT_TOKENS: u32 = 1000; pub(crate) const DEFAULT_DELAY: u32 = 1; /// Below this a routine step counts as long; above it the reminder never fires in practice. pub(crate) const TOKENS_RANGE: std::ops::RangeInclusive<u32> = 100..=200_000; /// A large delay would land the reminder long after the step it is about, so it is capp

    This code reference does not establish a working model identifier.

  3. · Catalogue listing · GitHub · xai-org/grok-build · code
    Excerpt
    e::app::ScreenMode::Fullscreen, current_title: None, } } #[test] fn split_model_rest_keeps_a_multi_word_label() { let mut state = ModelState::default(); let (id, info) = model_with_reasoning("grok-4.7", "Grok 4.7"); state.available.insert(id.clone(), info); assert_eq!( split_model_rest(&state, "Grok 4.7 Extra High") .map(|(model, token)| { (model.0.to_string(), token.to_string()) }), Some(("grok-4.7".to_string(), "Extra High".to_string())) ); assert_eq!( split_model_rest(&state, "Grok 4.7 high").map(|(_, token)| token), Some("high") ); assert!(split_model_rest(&state, "Grok 4.7").is_none()); assert_eq!( split_model_rest(&state, "grok-4.7 Extra High") .map(|(model, token)| (model.0.to

    This code reference does not establish a working model identifier.

  4. · Catalogue listing · Vertex AI · quotas
  5. · Alex's posts · Alex Getman
    Excerpt
    Grok 4.7 is #6 on Artificial Analysis
  6. · Alex's posts · Alex Getman
    Excerpt
    🚨 Grok 4.7 is out!
  7. · Alex's posts · Alex Getman
    Excerpt
    🚨 Grok 4.7 just appeared in Google Cloud That usually means the official release is very close.

Evidence links

  • · Catalogue listing · xAI API · grok-4.7

Alex's take

Signals 1

xAIReleased

Grok 4.7

Confirmed15 evidence links

Alex's posts 2

Specifications

Grok 4.7 is a model from xAI, released on 21 Sep 2026. The context window is 500,000 tokens. Pricing is $2.00 (Input ≤ 199,999 tokens); $4.00 (Input > 199,999 tokens) per million input tokens and $6.00 (Input ≤ 199,999 tokens); $12.00 (Input > 199,999 tokens) per million output tokens.

Context window
500,000
Input
image · text
Output
text
Input · USD / 1M tokens
$2.00 (Input ≤ 199,999 tokens); $4.00 (Input > 199,999 tokens)
Output · USD / 1M tokens
$6.00 (Input ≤ 199,999 tokens); $12.00 (Input > 199,999 tokens)
Cached input · USD / 1M tokens
$0.50 (Input ≤ 199,999 tokens); $1.00 (Input > 199,999 tokens)

Questions

What is the context window of Grok 4.7?

The context window is 500,000 tokens.

How much does Grok 4.7 cost?

Pricing is $2.00 (Input ≤ 199,999 tokens); $4.00 (Input > 199,999 tokens) per million input tokens and $6.00 (Input ≤ 199,999 tokens); $12.00 (Input > 199,999 tokens) per million output tokens. Cached input is $0.50 (Input ≤ 199,999 tokens); $1.00 (Input > 199,999 tokens) per million.

When was Grok 4.7 released?

On 21 Sep 2026.