xAI Ships Grok 4.7, Cuts Price to $1.60/$4.80 per 1M Tokens With 500K Context

xAI Releases Grok 4.7

xAI's Grok 4.7 is now listed on OpenRouter, succeeding Grok 4.6 as the company's flagship model for coding, agentic tasks, and knowledge work. The listing shows a 500,000-token context window and pricing of $1.60 per 1 million input tokens and $4.80 per 1 million output tokens — down from $2/$6 per 1M for Grok 4.6, a roughly 20% price cut on both input and output.

Notably, the OpenRouter model page attributes Grok 4.7 to "SpaceXAI" rather than xAI. xAI has not issued a separate public announcement confirming a rebrand, so this naming should be treated as unconfirmed pending official word from the company.

What's New, According to xAI

xAI claims Grok 4.7 improves on Grok 4.6 in three areas:

• Long-running software engineering tasks — the model is reportedly better at sustaining multi-step coding work and verifying its own output.
• Long-context handling — supported by the 500K token window.
• Professional knowledge work — including drafting documents and presentations.

According to xAI, the model was trained with an extended reinforcement learning run weighted toward problems that take multiple hours to solve, rather than short single-turn tasks. The company also says Grok 4.7 natively understands the "Grok Bot" harness used for conversational deployments, and ships with a new safety stack designed to resist jailbreak attempts while keeping refusal rates low for legitimate cybersecurity and biology-related queries.

xAI's reported benchmark results for Grok 4.7 use the "xhigh" reasoning effort setting — the top of a configurable reasoning-effort range. No specific benchmark scores (e.g., SWE-bench, MMLU, GPQA) were disclosed in the listing, so direct comparison to Grok 4.6 or competing models like Claude and GPT-5-class systems cannot be verified at this time.

Pricing and Availability

On OpenRouter, standard routing through xAI's own infrastructure shows a P50 latency of 0.99–1.49 seconds and throughput of 54–64 tokens per second, with cached input priced at $0.40 per 1M tokens. A priority tier doubles pricing to $3.20/$9.60 per 1M tokens for lower latency and higher throughput guarantees. Reported uptime over the trailing 24 hours was 94.18%. The listing shows a release date of September 21, 2026.

Grok 4.7 joins a growing lineup of xAI models on OpenRouter, including Grok 4.20 (2M context, $1.25/$2.50), Grok 4.3 (1M context, tiered pricing), and Grok Build 0.1, a dedicated coding-agent model with a 256K context window.

What This Means

Grok 4.7 is an incremental update rather than a architectural leap — same 500K context window as Grok 4.6, with the primary changes being price reduction and claimed gains in agentic and long-horizon task performance. The absence of published benchmark numbers means xAI's improvement claims can't yet be independently verified. The "SpaceXAI" branding on the OpenRouter listing is the most unusual detail here and warrants confirmation before being treated as an official corporate rename. Buyers evaluating Grok 4.7 for production coding or agentic workflows should wait for third-party benchmark results (e.g., SWE-bench Verified, LiveCodeBench) before switching from Grok 4.6 or competing models.