Musk Orders Tesla Staff to Use Grok 4.5 as $200 AI Cap Spares xAI
Updated
Updated · Electrek · Jul 10
Musk Orders Tesla Staff to Use Grok 4.5 as $200 AI Cap Spares xAI
2 articles · Updated · Electrek · Jul 10
Summary
$200 a week in new spending limits on outside AI tools was followed by Musk’s memo telling Tesla employees to use Grok 4.5 “when possible” and send him feedback directly.
Grok’s appeal is cost, not performance: Musk said Claude Fable 5 is “definitely better,” while Grok 4.5 costs about $0.13 per task versus $1.57 for Claude Fable 5.
Benchmark results show the trade-off clearly, with Grok 4.5 ranking 9th overall at 76.3, posting the lowest coding score on one leaderboard at 68.6, and scoring 53% on DeepSWE 1.1 versus Claude Fable 5’s 70%.
Tesla had already been testing Grok internally for months, but people familiar with usage said engineers broadly preferred Anthropic’s Claude for day-to-day development work.
The directive sharpens scrutiny of Tesla’s AI procurement because the cap hits Anthropic, OpenAI and Google tools while exempting Grok from Musk’s affiliated xAI business.