Elon Musk directs Tesla staff to use Grok AI model

Elon Musk told Tesla employees on July 10 to switch to the company‑owned Grok 4.5 model “when possible,” highlighting its $0.13 per‑task cost versus the $1.57 cost of competing models [1].

Cost‑driven directive

Tesla introduced a $200 weekly limit on employee use of third‑party AI tools such as Anthropic, OpenAI, and Google models earlier that week [1]. The cap excludes xAI’s Grok, which Musk promoted as a cheaper alternative. In the memo, he asked engineers to email him directly with feedback, signaling personal involvement in the rollout [1].

Performance versus price

Benchmark data shows Grok 4.5 trailing leading models: it placed ninth on a multi‑domain leaderboard with a 76.3 score, and its coding benchmark of 68.6 was the lowest among the top tier [1]. By contrast, Claude Fable 5 scored 70 on the same coding test [1]. Musk acknowledged the gap, noting that “most tasks don’t require Fable‑level capability,” but emphasized the cost advantage as the primary driver [1].

Internal preferences

Four engineers familiar with Tesla’s AI usage said the team still prefers Anthropic’s Claude for day‑to‑day development, despite the cost incentive [1]. The internal testing of Grok has been ongoing for months, with xAI product lead Andrew Milich assisting Tesla staff on troubleshooting [1].

What to watch

• Token‑cost comparison: Monitor any changes to Grok’s per‑task pricing versus competitors, which could shift the cost‑benefit calculus.
• Benchmark updates: New leaderboard scores for Grok 4.5 or future versions may affect internal adoption decisions.
• Employee feedback: Musk’s request for direct emails could surface internal sentiment and potential adjustments to the AI spend policy.

The move underscores a tension between cost savings and tool performance at a publicly traded firm, raising questions about how internal AI tool mandates might influence broader industry adoption patterns.