SpaceXAI released Grok 4.7 on Amazon Bedrock, saying it is trained to work longer on more difficult tasks and knows when to check its own work. It is the company's first major update to its foundation model series since the launch of Grok 4.6.
SpaceXAI reported a 37.6% score on multi-hour terminal work, measured on the Terminal-Bench 4.0 benchmark. That compares with a 20.3% score on Grok 4.6.
Grok 4.7 is built on a larger base model and targets tasks requiring sustained reasoning. Availability begins on Amazon Bedrock, initially for developers and enterprises using AWS services.
"Grok 4.7 uses a larger base model than Grok 4.6 and was trained with a longer reinforcement learning run on a harder mix of tasks," said SpaceXAI. The model is better at verifying its own work and managing longer context.
The announcement follows the launch of Grok 4.6, which introduced improvements in multi-turn reasoning and tool calling. SpaceXAI said the update reflects the company's focus on long-horizon tasks.
SpaceXAI did not say how the model performs on single-turn tasks, and the company raised the question of how to measure reasoning effort for different use cases. The source says the next step is to evaluate the model on tasks shaped like the user's own workload.
Source: awsml
