xAI has released Grok 4.7, the company's most capable model for coding and knowledge work to date, after weeks of public teasing and at least one delay. Announced on the company's site on Sunday, the model lands at the same price and serving speed as its predecessor Grok 4.6: $2 per million input tokens and $6 per million output tokens, roughly half the list price of OpenAI's GPT-5.6 Sol Max ($4/$20) and well under Anthropic's Fable 5.1 Max ($10/$50).

For the latest breaking AI news and analysis, visit AI Buzz Wire.

A Larger Base Model, Trained for Long Tasks

According to xAI's announcement, Grok 4.7 uses a new, larger base model compared with Grok 4.6 and was trained with a longer reinforcement learning run on a harder mix of tasks, weighted toward problems that take many hours to complete. The company says the model is better at verifying its own work and managing longer context, and that it was trained to natively understand the Grok Bot harness, improving conversational tasks and general knowledge work.

The release follows a noisy run-up. Elon Musk first said in early September that Grok 4.7 would land within ten days, then acknowledged in mid-September that the model needed a few more days of refinement, according to TeslaNorth.com. It is now shipping, and GitHub confirmed the same day that Grok 4.7 is available in GitHub Copilot.

The Benchmark Picture: Strong, Not Sweeping

The benchmark data published by xAI shows a model that leads in several long-horizon categories while remaining behind Anthropic's most expensive configuration on others:

  • CursorBench 4.0, which stresses longer-running coding tasks: Grok 4.7 scores 46.3%, ahead of GPT-5.6 Sol Max (41.7%) and Grok 4.6 High (40.4%), but behind Fable 5.1 Max (51.8%).
  • Terminal-Bench 4.0, multi-hour terminal work: 38.0%, up sharply from 20.3% for Grok 4.6 and just ahead of GPT-5.6 Sol Max (37.3%), though Fable 5.1 Max leads at 57.9%.
  • EEBench, an electrical engineering benchmark: 64.0%, a large jump from Grok 4.6's 53.0% and the highest of the models listed.
  • DeepSWE v1.1: 71.0% at high effort, versus 65.2% for Grok 4.6 and 72.7% for GPT-5.6 Sol Max.
  • AA Briefcase v1.1, multi-hour office work: 1,657, up from 1,546, and close behind Fable 5.1 Max at 1,678.
  • Harvey Legal Agent Benchmark: 19.6%, ahead of Grok 4.6 (15.8%) and far ahead of GPT-5.6 Sol Max (2.5%) and Fable 5.1 Max (6.7%) on this measure.
  • HealthBench Professional: 56.7%, improving on Grok 4.6's 48.5% but trailing GPT-5.6 Sol Max (60.5%) and Fable 5.1 Max (62.1%).

On GDPval, a professional knowledge-work benchmark scored by Elo, xAI's chart puts Grok 4.7 at xhigh reasoning on 1,695 — up from 1,605 for Grok 4.6, behind Fable 5.1 Max (1,735) and ahead of GPT-6 Astra Max (1,542).

Price Is the Story

The most aggressive claim in the announcement is economic rather than technical: xAI describes Grok 4.7 as twice as fast at half the price of comparable models, and the published pricing table backs the headline comparison against the top-tier configurations of its two chief rivals. Grok 4.7's $2/$6 token pricing is unchanged from Grok 4.6, meaning buyers get the generation-over-generation improvement without a price increase.

That positioning extends a strategy xAI has pursued across the Grok 4.x line: match or approach frontier quality while undercutting the premium pricing tiers of OpenAI and Anthropic, then distribute aggressively through partners including GitHub, Cursor, and OpenRouter.

What to Watch

The honest caveat is that xAI published the comparison chart itself, and independent verification from benchmarking aggregators will take days. On the numbers xAI chose to highlight, Grok 4.7 is a genuine step up from Grok 4.6 — most visibly on Terminal-Bench 4.0 and EEBench — but it does not sweep Fable 5.1 Max, which retains the lead on coding and health benchmarks while costing five times more per output token.

For developers running long, multi-hour agentic workloads, the price-performance math may matter more than raw leaderboard position. Grok 4.7 is available now through xAI's API and in GitHub Copilot.

Stay Ahead of AI

For continuous coverage of the AI industry's most important developments, bookmark AI Buzz Wire and never miss a breaking story.

Read more AI news