xAI · Grok
Grok 4.7
xAI's current flagship for coding and knowledge work, released 21 September 2026 on a new, larger base model with a longer reinforcement-learning run weighted toward problems that take hours to complete; xAI says it is better at verifying its own work and managing longer context, at the same headline price as Grok 4.6.
Start here
- What this helps you do
- Decide whether Grok 4.7 is the right model for a job you have, or one to leave alone.
- What you leave with
- A plain read of what Grok 4.7 is for, what to avoid giving it, and the facts behind that.
- What to check before trusting it
- Released 21 September 2026, so real-world track record is minimal at review time.
Treat this as a migration decision from 4.6, not a new default. Test the longer-running and self-verification claims on your own task before moving anything in production.
Use it for
- Longer-running coding tasks that benefit from more self-verification
- Professional knowledge work such as legal analysis or engineering review
- Office-based professional tasks where Grok 4.6 was the prior default
Being the newest flagship is not itself a reason to migrate a working Grok 4.6 workflow; access, headline pricing and the context window are unchanged, so the case has to come from the deeper reasoning and self-verification claims holding on your own task, not the provider's framing alone.
Effort without the jargon
The task is genuinely long-running or difficult enough that the longer reasoning pass and extra self-checking change the outcome.
Advanced recordCost, limits, privacy and operations
Grok 4.7 keeps 4.6's headline input, output and cached-input pricing and context window exactly; the entire upgrade case rests on the longer reasoning pass and self-verification claims, so use the published table as a planning input, then measure whether that extra reasoning actually pays on your task.
- API cost
- Officially disclosedxAI prices Grok 4.7 in the same two context tiers as 4.6, unchanged: US$2 input and US$6 output per million tokens below 200,000 tokens, doubling to US$4 and US$12 at or above it, with cached input at US$0.50 below 200,000 tokens and US$1.00 at or above it. A faster Grok 4.7 variant is available at double the price. Tool and other service charges can be separate.
- Context window
- Officially disclosed500,000 tokens
- Maximum output
- Not disclosedNot disclosed in the cited record
- Knowledge cutoff
- Officially disclosedMay 2026
- Recorded inputs
- Not enumerated in cited record
- Recorded output
- Text
Data handling is a product decision
Officially disclosedxAI says API inputs and outputs are not used for training without permission and are retained for 30 days by default for abuse audit. Team-wide ZDR removes storage but disables stateful features.
Operational constraints
- Released 21 September 2026, so real-world track record is minimal at review time.
- Current events require search tools; the base model has no live-event access without them.
- Zero Data Retention disables stateful Responses, Files, Collections, Batch and deferred completions.
- The stable and latest aliases can move; use a dated version where workflow consistency requires it.
- The cited record does not enumerate image or other non-text input support for this model, unlike the explicit image note on Grok 4.5.
Record details
- Evidence readiness
- informational
- Guidance confidence
- moderate
- Evidence revision
- grok-47@v1
Unknown means the cited official record does not disclose a safe value. It is not an estimate. Prices are provider list prices in USD where stated and can change before this record's review date.
Keep the evidence separate
xAI calls Grok 4.7 our most capable model for coding and knowledge work, built on a new, larger base model with a longer reinforcement learning run on a harder mix of tasks weighted toward problems that take many hours to complete, and says it is better at verifying its own work and managing longer context than Grok 4.6.
No sufficiently useful independent note has been added yet. That absence is not filled with our guess.
No Human Bit result is claimed; use this as evaluation guidance only.
Not scheduled
No Human Bit result is claimed for this model yet.