Grok 4.7 nearly doubles its predecessor on long-running coding and lands on Amazon Bedrock

Grok 4.7 is xAI's coding-focused upgrade. Reports say it nearly doubles Grok 4's score on a long-running Terminal-Bench-style coding benchmark and offers a 500,000-token context window. The company now operates as SpaceXAI. It also introduced bundles combining Grok, X and Cursor at $30, $100 and $200 per month, with 4x, 15x and 50x Plus-level usage.

AWS made Grok 4.7 available on Bedrock the same day. In a detailed post, AWS built a tool-calling agent on Grok 4.7 and ran it on three reasoning tasks at four reasoning-effort levels, measuring correctness against token cost per task. The guidance is practical: tune effort per request, because maximum reasoning often adds cost without improving accuracy on simpler steps.

Grok 4.7 competes in a crowded week for coding models. Gemini 4 Argon tops the Vals Index, Meta's Muse Wonder targets agentic coding, and Mistral Large 4 claims open-model leads on DeepSWE. Being on Bedrock gives xAI a far larger enterprise distribution channel than its own API.

Community debate centers on the bundles. Developers ask whether tying Cursor to Grok and X subscriptions at $100–$200 creates tool lock-in, and whether the benchmark gains hold up in independent testing.