Breaking

SpaceXAI Releases Grok 4.7 for Coding and Knowledge Work With Lower-Cost Frontier Benchmarks

September 21, 2026 at 12:45 EDT

Foundation Models
AI Agents
Software Dev & Coding

SpaceXAI released Grok 4.7 on September 21, positioning the model as its most capable system for coding, agentic tasks and professional knowledge work. The company says the new model is available in Cursor, Grok Build, the Grok API, third-party coding harnesses, model routers and cloud platforms, with API pricing starting at $2 per million input tokens and $6 per million output tokens.

Sep. 21, 2026 · SpaceXAI

Grok 4.7 pushes coding agents on price-performance

The new model is aimed at coding, agentic tasks and knowledge work, with a 500,000-token context window and the same base API price as Grok 4.6.

46.3%

CursorBench 4.0

38.0%

Terminal-Bench 4.0

500K

Context window

$2/$6

Input/output per 1M tokens

Terminal work nearly doubles from Grok 4.6

A proportional stacked-block view of Terminal-Bench 4.0 scores.

38.0%Grok 4.7

20.3%Grok 4.6

Takeaway: the published Terminal-Bench 4.0 score rises from 20.3% to 38.0%, while pricing stays at $2 input and $6 output per 1M tokens.

Why builders care

The model is available in Cursor, Grok Build, API workflows and coding harnesses, making same-price upgrades practical for existing Grok 4.6 users.

What remains open

The published table is not a clean sweep: Fable 5.1 leads several listed benchmarks, and real-world adoption will depend on independent developer testing.

Release shape

ModelLarger base model and longer reinforcement-learning run

→

Use casesCoding, agentic tasks and professional knowledge work

→

AccessCursor, Grok Build, Grok API and model platforms

Continue reading

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