xAI launched Grok 4.5 (now under SpaceXAI) with aggressive “beats Fable” claims from Elon Musk.
Independent benchmarks rank it lower than Musk suggests, but its cost per task is far cheaper.
Musk conceded Fable is better, then told Tesla staff to switch to Grok anyway.
Few AI launches come with this much noise. Grok 4.5 arrived, Elon Musk declared victory across benchmarks, and independent testers immediately pushed back. The truth sits somewhere in the middle, and the real story is price.
The claims versus the benchmarks
Musk went straight at the market leader. “Grok 4.5 even ranks slightly above Fable, which is an incredibly good model, on some software benchmarks!” he posted on July 12. He also claimed a top spot on the SWE marathon, a long-horizon coding benchmark.
See the post here:
https://x.com/elonmusk/status/2076434607562531049
Independent data tells a more modest story. On an aggregate multi-domain leaderboard, Grok 4.5 lands 9th overall at 76.3, behind multiple models from OpenAI, Anthropic, and Google, and its coding score of 68.6 is the lowest of any model on the board. Artificial Analysis ranked it fourth on its Intelligence Index, behind Claude Fable 5, GPT-5.5, and Claude Opus 4.8.
Why it still matters: the price
Grok's pitch is not “best,” it is “cheap enough.” Grok 4.5 runs at roughly $0.13 per task on that leaderboard, versus $1.57 for Claude Fable 5. For high-volume agentic work, that math changes budgets fast.
Even Musk admitted the quality gap. “In fairness, Fable is definitely better than Grok 4.5, but most tasks don't require Fable-level capability,” he wrote on X.
Then things got awkward internally. Musk told Tesla staff to move to Grok “when possible” given its lower token costs, days after Tesla capped employee spending on third-party AI tools, a cap that applies to Anthropic, OpenAI, and Google but pointedly exempts Grok.
The honest takeaway for buyers: Grok 4.5 is a genuine value play, not a frontier king. If your workload is heavy and simple, the cost savings are real. If you need top-tier reasoning or coding, the leaderboards still point elsewhere. Choose on the task, not the headline.
Quick Links:
