xAI ships Grok 4.7: big coding gains at $2/$6 pricing, one-fifth of Fable 5.1
AGI Hunt · wechat · 2026-09-22
xAI released Grok 4.7, positioned as its strongest coding and knowledge-work model at unchanged pricing ($2/$6 per million tokens, 1/5 of Fable 5.1).
Benchmarks
- CursorBench 4.0 (xhigh): 46.3%, up from 40.4%, beating GPT-5.6Sol's 41.7% but trailing Fable 5.1's 51.8%
- Terminal-Bench 4.0: nearly doubled from 20.3% to 38.0%
- DeepSWE v1.1: 71.0%, above Fable 5.1, below GPT-5.6Sol's 72.7%
- Vertical wins: Harvey legal agent benchmark 19.6% vs 2.5%/6.7%; EEBench 64.0%; AABriefcase 1657, near Fable 5.1; HealthBench Professional (56.7%) is the weak spot
Under the hood
- New, larger base model with longer RL training on multi-hour tasks; stronger self-verification and long-context management; native GrokBot harness understanding
Other
- Demo building an open-world city game shows a clear jump over 4.6; four reasoning levels in GrokBuild; fast variant doubles speed and price
- Safety: best-calibrated version yet, 62.4% on LatchBio biosafety, only 3.3% high-risk cybersecurity prompts allowed
Overall strategy: 90% of frontier performance at a fraction of the price.
More from coding & agent
- How to cut Fable 5.2 token costs by 90%: Arc routing plus a lean CLAUDE.md — EXM7777 · 2026-11-12
- Hamel Husain's complete AI evals framework for PMs and engineers — aakashgupta · 2026-11-12
- AI coding in one year: from unusable to full automation with no code review — angelaqdai · 2026-11-12
- Interactions Tool Searches Web Animations by Motion for Claude Code to Rebuild — yuwen_lu_ · 2026-11-12
- Agent Substrate Calculator Estimates Session Resume Speed Under Load — pjausovec · 2026-11-12
- Senate Bill Would Force Big Platforms to Accept AI Agents, Raising Auth Privacy Questions — Prof_CarlaReyes · 2026-11-12