Breaking

SpaceXAI Launches Grok 4.7 for Coding and Agentic Work at Grok 4.6 Prices

September 21, 2026 at 17:16 EDT

Foundation Models
AI Agents
Software Dev & Coding

SpaceXAI released Grok 4.7 on September 21, positioning the model as an upgrade for coding, agentic tasks and knowledge work while keeping the standard token rates level with Grok 4.6. The company says the new model uses a larger base model, longer reinforcement-learning training on harder multi-hour tasks, stronger self-verification and improved long-context handling.

Sep. 21, 2026 · SpaceXAI

Grok 4.7 moves into longer agent work

The new model keeps Grok 4.6's standard token rates while improving coding and knowledge-work benchmarks. The main question is whether stronger task completion offsets heavier token use.

46.3%

CursorBench 4.0 score

$2 / $6

Input and output per 1M tokens

500k

Maximum context window

Coding Agent Index

Output-token trade-off

Approximate output tokens per Intelligence Index task.

36k

Grok 4.6 high

81k

Grok 4.7 xhigh

The upgrade improves agentic results, but output-token use can more than double on some measured tasks.

What improves

Better results on long-running coding, multi-hour knowledge work and richer open-ended build prompts.

What to watch

Higher token consumption and longer task times can change the real cost of production workloads.

Rollout path

xAI API→Cursor→Grok Build→GitHub Copilot

Continue reading

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