GMAsia
    AI News·21 Sept 2026·via Testing Catalog·Covered by 4 sources

    SpaceXAI releases Grok 4.7 for coding and knowledge work

    SpaceXAI has released Grok 4.7, its newest model for coding and knowledge work. The model is accessible via Grok Build, its API, Cursor, and various cloud platforms. Grok 4.7 shows performance gains, scoring 46.3% on CursorBench 4.0 and 71.0% on DeepSWE v1.1. Pricing remains at $2 per million input tokens and $6 per million output tokens, with a faster option available. The release also includes an enhanced safeguard stack for improved refusal and jailbreak resistance.

    Nexa's Summary

    SpaceXAI's Grok 4.7 launch is a direct challenge to other frontier models, not just an incremental update. The company is pushing for market share by maintaining its $2 per million input token price and doubling output speed for a premium. Its 46.3% score on CursorBench 4.0 and 71.0% on DeepSWE v1.1 position it as a serious contender for development teams.

    This release will pressure Asian AI developers and platform providers. Companies like Alibaba Cloud and Tencent Cloud, which offer access to various models, must now consider integrating Grok 4.7 to remain competitive. Developers in markets like Singapore and South Korea, focused on enterprise AI solutions, will evaluate Grok 4.7's longer task execution and self-checking features against existing options.

    The key thing to watch is Grok 4.7's adoption rate among Asian coding teams within the next six months. Its 3.3% risky prompt allowance on HackerBench v0.3 suggests strong safety. If this translates to real-world trust and efficiency, it could shift model preferences away from other established players in the region.

    #spacexai
    Go deeper
    Original reporting by Testing CatalogWe don't republish, read the full story →

    Related reading

    6 stories