Elon Musk has hit back at a report which claimed that the tech billionaire has told the Tesla staff to switch to Grok 4.5, a model he admits is worse. He has clarified that he does not force his employees to use his own software blindly if a competitor’s product performs better.The clarification comes after an internal memo leaked (via The Information) from Tesla last week, showing that the electric vehicle maker had placed a strict $200 weekly cap on employee spending for third-party AI tools like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude.Taking to his social media platform, X, Musk posted: “I just asked Tesla & SpaceX to try out Grok 4.5 to see if it solves their task, not use it no matter what! They should continue to use other AI models if those models outperform Grok.”Recently, reports said that companies have told staff to use the in-house models. Most recently, Microsoft is said to have started to migrate to indigenously developed AI models rather than those developed by OpenAI.
The battle of the benchmarks
The leaked memo, originally reported by The Information, instructed staff to transition their daily workloads to Grok “when possible” and urged engineers to email Musk directly with performance feedback. However, multiple sources close to internal testing revealed that several Tesla engineers strongly prefer Anthropic’s Claude for day-to-day software programming.Independent data claims that Grok 4.5 currently ranks 9th overall, trailing behind several premier models from OpenAI, Google, and Anthropic. On coding-specific benchmarks, Grok 4.5 scored a 53% success rate, significantly below Claude Fable 5’s 70% rating.Musk has openly acknowledged the current performance gap between Grok and its rivals, but argued that the decision ultimately comes down to corporate cost-efficiency.“In fairness, [Claude] Fable is definitely better than Grok 4.5, but most tasks don’t require Fable-level capability,” Musk noted.The primary advantage of Grok 4.5 is its radically lower operating cost. Running a standardized task on Grok costs roughly $0.13, compared to a steep $1.57 on Anthropic’s top-tier model.“Our token efficiency seems to be better than any other AI model and we see many more ways to improve inference per watt, so I think Grok will continue to be the best value for money of the frontier models,” he explained.
