[DigitalToday reporter Jinju Hong] Elon Musk instructed Tesla employees to switch to xAI's artificial intelligence (AI) model Grok where possible. Tesla set a weekly cap on budgets for using external AI tools, while making Grok an exception, as it reshapes its internal development environment around an affiliated model.

On July 10 local time, electric vehicle outlet Electrek reported that Tesla earlier that week limited employees' budgets for AI tools from Anthropic, OpenAI and Google to $200 per person per week. xAI's Grok was excluded from the cap.

In an internal memo to employees, Musk ordered them to switch to Grok where possible, saying Grok 4.5's token costs are lower than rival models. He also asked engineers to use the model directly and email him problems and suggestions for improvement.

The move came as Tesla has tested a beta version of Grok internally for months. xAI head of product Andrew Milich was reported to have worked with Tesla employees on fixing issues and improving features. But people familiar with internal use said Tesla engineers have preferred Anthropic's Claude for actual development work.

Musk's order also ties into controversy over Grok 4.5's performance. xAI unveiled Grok 4.5 on July 9, and Musk introduced it on X as an "Opus-class" model. But on a consolidated benchmark cited by Electrek, Grok 4.5 ranked ninth overall with a score of 76.3. It ranked below multiple models from OpenAI, Anthropic and Google, and its coding score was 68.6, the lowest among those compared.

A similar result appeared in the DeepSWE 1.1 benchmark, which evaluates the ability to resolve GitHub issues. Grok 4.5 scored 53 percent, while Claude Fable 5 scored 70 percent. During the release process, questions were also raised about benchmark reliability. AI coding tool company Cursor said a previous snapshot of its codebase was mistakenly included in Grok 4.5's training data. Related comparison metrics were excluded from public materials as a result.

Musk also did not fully deny a performance gap between Grok 4.5 and rival models. He wrote on X, "To be fair, Fable is clearly better than Grok 4.5." He added that "most tasks do not require performance at the level of Fable," stressing cost effectiveness.

The biggest strength of Grok that Musk highlighted was price. According to the outlet, Grok 4.5 costs about $0.13 per task, far lower than about $1.57 for Claude Fable 5. Tesla's decision to cap spending only on external models while exempting Grok has been interpreted as linked to that cost structure.

But because employees have used Claude more when choosing AI tools on their own, the order carries meaning beyond simple cost cuts. It is seen as strongly aimed at shifting the company's internal AI use toward Grok.

xAI was also introduced as currently being incorporated into SpaceX. That structure has also shown that Tesla's choice of AI tools is directly linked to adopting products from other companies involving Musk. The focus going forward is on whether Tesla engineers' actual usage patterns will shift toward Grok, and what balance will be found between cost savings and development productivity.