SpaceXAI's New Grok 4.7 Improves on Coding, Maintains Cheaper Token Rates than Peers
The model improved on long-running coding benchmarks and starts at $2 per million input tokens, SpaceXAI said.
- On Sunday, SpaceXAI released Grok 4.7, pricing the model at $2 per million input tokens and $6 per million output tokens with a 500,000-token context window.
- SpaceXAI paired its base model with extended reinforcement learning runs deliberately weighted toward harder tasks, including problems that take "many hours" to complete, targeting self-verification and long-context management.
- Endurance benchmarks show Grok 4.7 scored 38.0% on Terminal-Bench 4.0, up from 20.3% for Grok 4.6, while improving from 40.4% to 46.3% on CursorBench 4.0 and reaching 1,657 on Briefcase.
- Despite gains, Grok 4.7 trails Anthropic's Claude Fable 5.1, which scores 57.9% on Terminal-Bench 4.0; benchmarks show even top-performing models failed more than 60% of the time on private codebases.
- SpaceXAI trained the model to natively understand the Grok Bot harness, mirroring OpenAI's recent decision to open its Codex harness as the Agents API for managed long-running agent services.
42 Articles
42 Articles
SpaceXAI presented a coding model with improved self-checking and a long context.
Ilona Maska xAI introduced the updated Grok 4.7 neural network, which is characterized by improved programming skills and enhanced security in complex intellectual tasks. "Grok 4.7 is our most functional model for programming and intellectual work. It works longer on complex tasks, more carefully test its performance and is equipped with our best security mechanisms to date," says the company's official release, delivers RIA News.
SpaceX AI Released Grok 4.7, Its Most Advanced Model for Coding and Knowledge Work
New York SpaceXAI released Grok 4.7, its most advanced model for coding and knowledge work, and according to the company, it is “highly competitive in its class.” It has been released at the same price and speed as Grok 4.6. However, Grok 4.7 uses a new, larger base model compared to Grok 4.6. SpaceX explained that “it was trained with a longer reinforcement learning run on a harder mix of tasks, weighted toward problems that take many hours to …
SpaceXAI announced Grok 4.7 on September 21, 2026. It offers significant improvements over Grok 4.6 at the same price and speed, with CEO Elon Musk touting it as "a powerful combination of intelligence, speed, and low cost." Read more...
Coverage Details
Bias Distribution
- 50% of the sources are Center, 50% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
