Breaking

SpaceXAI Releases Grok 4.7 for Coding, Agentic Work and Long-Context Tasks

September 22, 2026 at 21:16 EDT

Foundation Models
AI Agents
Software Dev & Coding

SpaceXAI has released Grok 4.7, positioning the new model as its strongest system for coding, agentic workflows and knowledge work, with public availability through Cursor, Grok Build and the xAI API. The launch also drew attention from developers testing whether the model can turn short prompts into working interactive applications, including a one-prompt 3D jet-engine visualizer demo.

Sep. 23, 2026 · SpaceXAI

Grok 4.7 shifts toward longer agent work

The new model is aimed at coding, self-checking and multi-hour knowledge tasks, with early demos highlighting richer interactive web apps from compact prompts.

500k

maximum context window

$2/$6

base input and output price per 1M tokens

reasoning effort levels: low to xhigh

CursorBench 4.0: measured gain over Grok 4.6

40.4%Grok 4.6

46.3%Grok 4.7

Each block represents about five percentage points, showing a modest but visible increase on a longer-running coding benchmark.

What changed

Larger base

→

Longer RL run

→

Stronger self-checking

Early upside

Developers are using it for richer 3D and interactive web-app prototypes with fewer follow-up prompts.

Practical limit

Longer thinking may consume more session budget, so task-level cost is still under scrutiny.

Continue reading

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