Model update · Version comparison
Grok 4.7: compare longer tasks with the model you use
The release emphasizes longer work and checking results, with standard API rates unchanged from 4.6. Compare access and speed variants before testing a real task.
The announcement lists Cursor, Grok Build and API access; it does not establish identical access or allowances for every chat subscription.
The focus is sustained work
The September 21 announcement describes a larger base model than 4.6, with training weighted toward tasks that take hours, stronger self-checking and longer-context management. Documents and presentations are also highlighted.
A practical question is whether a complex assignment needs fewer interruptions, retains its requirements and produces a complete result. Published scores are useful leads, but remain vendor-reported evidence rather than proof for your own work.
Sources and further reading
- Grok 4.7 announcementOfficial · September 21Version changes, vendor benchmarks, access and API rates; not our independent evaluation.
Standard rates stay the same; fast is separate
Standard API rates start at $2 per million input tokens and $6 per million output tokens, unchanged from 4.6. A fast variant doubles output speed and price. Total task cost still depends on input, output and the number of calls.
If you work in Cursor or another tool, first inspect its model selector and billing. API rates, tool subscriptions and chat allowances are distinct; unchanged model rates do not guarantee unchanged cost in every product.
Sources and further reading
- Grok 4.7 announcementOfficial · September 21Version changes, vendor benchmarks, access and API rates; not our independent evaluation.
Use the same assignment as your reference
Choose work you can evaluate: turn the same material into a report or add one feature to a small project. Give Grok and your current model the same inputs, requirements and output format, then compare omissions, revisions and usable results.
This is an editorial comparison method, not a completed test. Keep using a model that handles your short tasks well; trying the newcomer on a longer task that often stalls may reveal where it fits.
Sources and further reading
- Grok 4.7 announcementOfficial · September 21Version changes, vendor benchmarks, access and API rates; not our independent evaluation.