xAI Launches Grok 4.7 With 2.1T Parameters, Trails Claude and OpenAI
In brief
- Grok 4.7 launched Monday with 2.1 trillion parameters, 40% more than Grok 4.6
- Model scores 1695 on GDPval; Claude Fable 5.1 leads at 1735
- Available immediately in Grok app, Cursor, Grok Build, and xAI API
- Pricing: $2 per million input tokens, $6 per million output tokens
- Musk delayed launch at least five times since late July
Benchmarks Show Gains, But Not the Lead
Grok 4.7 scored 1695 on GDPval, while Claude Fable 5.1 topped the chart at 1735. On AA-Briefcase, Grok 4.7 posted 1657 against Fable 5.1's 1678. The model lands in the middle on CursorBench 4.0—pricier per task than GPT-6 Astra and Claude Sonnet 5, but short of Fable 5.1. It's a step forward from Grok 4.5, which launched in July with third-place scores behind Claude and OpenAI's models, but not yet a frontier leader.
xAI says the model spends longer working through hard problems and double-checks its own answers more often than Grok 4.6. The model costs $2 per million input tokens and $6 per million output tokens—competitive pricing for a mid-tier performer.
Training Data and Roadmap
xAI folded supplemental training data from SpaceX into Grok 4.7, including Starlink satellite telemetry, manufacturing records, and engineering failure logs. That integration sets it apart from purely internet-trained models, though it doesn't yet close the gap on Claude or OpenAI's latest releases.
Musk had walked back the timeline at least five times since late July, with promises ranging from "four weeks out" to "a few days." The repeated delays underscore the pressure on xAI to compete in a market where frontier models release every few months. Musk had already set expectations lower days before launch, writing that Grok 4.7 should land "roughly on par with" Anthropic's Claude Opus 5.0, not the newer Opus 5.1.
Looking ahead, Musk sketched plans for Grok 4.8 as a meaningful step up, Grok 4.9 in "Astra/Fable class," and Grok 5 as a possible frontier leader. For now, Grok 4.7 is a solid incremental release—faster, cheaper, and more capable than its predecessor, but still playing catch-up in a field where Claude and OpenAI set the pace.
