Summary
Matthew Berman breaks down xAI’s newly released Grok 4.7, testing Elon Musk’s own prediction that the model would land roughly on par with Anthropic’s Opus 5. Using xAI’s official blog post and benchmark data, Berman walks through CursorBench 4.0 results plotting performance against cost per task, showing Grok 4.7 trailing Opus 5 slightly at maximum thinking effort but running at roughly half the price.
The video compares Grok 4.7 against Opus 5, Fable 5.1, GPT 5.6 Sol, and GPT-6 Astra across multiple benchmarks, including Terminal Bench 4.0 (where Grok scored 38% versus Fable 5.1’s 57.9% and Astra’s 58.2%), a legal-work benchmark where Grok 4.7 led at 19.6% while a lesser-known model, Muse Spark 1.2, actually topped the field at 42%, and pricing at $2 per million input tokens and $6 per million output tokens — less than half the cost of GPT 5.6 Sol and roughly a fifth the cost of Fable 5.1 or GPT-6 Astra.
Berman argues that while Grok 4.7 isn’t the absolute frontier model, its low pricing stems from xAI’s compute overcapacity, making it a strong option for teams that prioritize cost-efficient automation over marginal accuracy gains. The video also includes a Zapier-sponsored segment on integrating Grok 4.7 into automated workflows.
📺 Source: Matthew Berman · Published September 22, 2026
🏷️ Format: Benchmark Test
