The Information Machine
Following·Day 6·first covered 18 Sep 2026·4 sources

Grok 4.7 launches across APIs, Copilot, Tesla, and enterprise tools

The gist

Grok 4.7 is being deployed across developer tools, enterprise software, and consumer hardware simultaneously, putting it in front of a wide range of users and use cases. The insurance claims demonstration and EEBench results show the model being positioned for domain-specific professional work alongside general coding tasks.

The full picture

xAI released Grok 4.7, built on a larger base model than its predecessor and trained with a longer reinforcement learning run on harder, longer-horizon tasks. The model is available via API at $2 per million input tokens and $6 per million output tokens, with a fast variant at double the speed and price. It is available in Cursor and Grok Build, rolling out gradually in GitHub Copilot for agentic coding and multistep workflows, and is now accessible in Tesla vehicles via Grok Bot. Copilot Enterprise and Business administrators control access via model policy settings, with new models automatically enabled under default settings unless overridden.

On xAI's EEBench, Grok 4.7 scored 64.0%, up 11 points from Grok 4.6's 53.0%, ahead of Fable 5.1 at 56.4% and GPT-5.6 Sol at 39.4%. The Neuron Daily attributed part of the electrical engineering strength to SpaceX's engineering expertise and noted that model selection is becoming domain-specific.

Box previewed a Grok 4.7-powered agent in Box AI Studio for insurance claims processing. In a demonstration on a $2 million claim, the agent caught an $82,000 duplicate invoice and a missing $64,000 supplier credit, and correctly interpreted policy language to determine that the Business Income waiting period did not apply to Extra Expense coverage. A deductible error in the test claim shifted the calculation by $143,000. The agent outputs a cited claims review for human adjusters; final coverage and payment decisions remain with the insurer.

For cybersecurity, xAI said Grok 4.7 scores highest on HackerBench v0.3, allowing only 3.3% of risky dual-use prompts through while rarely blocking legitimate security work, and is providing select cybersecurity partners invite-only access for defense research. The model was trained to natively understand the Grok Bot harness, which xAI said improves conversational and general knowledge performance.

Separately, a creator and their studio used Grok's image and video generation tools to produce a cinematic pilot episode in 9 days, generating 3,627 images and 3,043 videos at a total generation cost of $2,677.

How it developed
23 September 2026

Grok Bot available in Tesla vehicles for hands-free productivity tasks

22 September 2026

Grok 4.7 reported to score 64.0% on xAI's EEBench, up from Grok 4.6's 53.0%

21 September 2026

Box previews Grok 4.7-powered insurance claims agent in Box AI Studio

18 September 2026

Creator and studio produce cinematic pilot in 9 days using Grok image and video generation, generating 3,627 images and 3,043 videos for $2,677

Sources
The daily email

Want this in your inbox?

I send one email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free