Grok 4.5 Rebranded as SpaceXAI’s Enterprise AI, Now Touts Legal and Excel Capabilities After Past Controversies

ai and ml

The newly renamed SpaceXAI wants you to believe little ol’ Grok is all grown up

Elon Musk’s AI venture has a documented history of producing controversial models. Its earlier version, Grok, was linked to offensive impersonations such as “MechaHitler” and generated deep‑fake adult content, prompting regional bans.

SpaceXAI now claims that Grok 4.5 has been rehabilitated. According to the company, the model has purged its problematic histories and refocused on professional tasks, including legal analysis, spreadsheet modeling, and code generation.

“Today we’re launching Grok 4.5, SpaceXAI’s smartest model built to excel at coding, agentic tasks, and knowledge work,” the firm wrote in a blog post. “The model is equally adept at office work, scoring number one on Harvey’s Legal Agent Benchmark.”

SpaceXAI portrays the new Grok as a more office‑savvy assistant, shifting from youthful mischief to productivity. The firm emphasizes that the model can now handle complex professional responsibilities.

“Grok Build is capable of building complex Excel models that involve research from the web, multi‑sheet formula use, and even leaves stickies or notes behind for future reference,” the company notes.

While the model’s legal benchmark results are highlighted, the article also references past regulatory issues, noting that similar controversies have resulted in modest settlements for Musk’s companies. This context is offered as a reminder of the model’s potential legal utility.

SpaceXAI reports that Grok 4.5 was trained on tens of thousands of Nvidia GB300 GPUs, alongside Cursor, a platform the company is acquiring for $60 billion. The training process placed a strong emphasis on data quality, using deduplication, scoring, and domain‑focused selection to maintain high‑signal content.

Further refinement was achieved through reinforcement learning, the same technique used by major AI labs to instill chain‑of‑thought reasoning. This approach has reduced the number of thinking tokens required for complex problems and enabled inference speeds of up to 80 tokens per second.

Benchmark evaluations from Artificial Analysis indicate that Grok 4.5 performs comparably to OpenAI’s GPT‑5.5 and Claude Opus 4.8/Sonnet 5, though it still lags behind Anthropic’s Claude Fable. The model remains in preview, with full leader‑board integration pending.

Pricing is positioned as competitive: $2 per million input tokens and $6 per million output tokens, significantly lower than GPT‑5.5’s $5/$30 rates. Grok 4.5 will be available starting Wednesday through Grok Build, Cursor, and the SpaceXAI console, excluding users in the European Union, with EU rollout slated for mid‑July.

Source link