Musk calls Grok 4.7 "a solid workhorse" but admits it trails Anthropic's Opus 5.5
In an interview with China Media Group, Elon Musk conceded that Grok 4.7 is not as good as Anthropic's Opus 5.5, and said his AI company probably won't catch the frontier until next year.
Elon Musk is rarely the one to say his product is not the best. In an interview with China Media Group recorded at Tesla's engineering headquarters, he did exactly that.
Discussing the latest release from his AI company, Musk called Grok 4.7 "a solid workhorse of a model," then added: "It's not as good as say, um, Opus 5.5 that just got released." When the interviewer noted the rare humility, Musk's explanation was arithmetic: "we've only been around for about 3 years doing AI whereas Anthropic's been around for six years." The full transcript of the interview has circulated widely since it aired.
Anthropic released Claude Opus 5.5 on September 22, and it has quickly become the model other labs are measured against. Musk told CMG he expects his team to "probably catch up to Frontier sometime next year most likely."
That is a notable climb-down from the summer, when Musk said Grok 4.5's internal evaluations put it "close to, perhaps exceeding Opus" and described it as "a solid workhorse in the same league as Opus," as reported at the time. Then it was framed as a near tie. Now, two releases later, Grok 4.7 is publicly conceded to be behind.
The real-world engineering pitch
Musk's answer to the gap is a strategy. He argued that Anthropic has built an AI "excellent at software engineering," while nobody has yet built one "excellent at real world engineering." His bet is that Grok can become that model by training on data from his other companies, SpaceX and Tesla: rocket telemetry, factory floors, vehicle fleets.
When the interviewer called that data a "secret weapon," Musk pushed back: "only a little bit so far."
It is a coherent thesis. Nobody else has a rocket company and a car company feeding training data into the same lab. Whether physical-world data compensates for a software-engineering deficit is the open question, and Musk's own timeline, next year, gives everyone a date to check his work against.
What else he claimed
Musk said Grok Bot, the company's agent offering, is "kind of like a personal digital assistant" and is growing "about 100 percent a month." That figure comes from Musk alone and has not been independently verified.
He also praised Chinese AI labs as "by far the best in terms of performance per unit of compute" and predicted China resolves its chip-manufacturing constraints within two to three years, according to coverage of the interview.
Why it matters
Founder concessions like this are rare precisely because they are priced in. When the person with the most to gain from hype tells you his model is second best, the ranking becomes harder to dismiss as partisan. For developers choosing a coding model, the practical read is straightforward: Opus 5.5 remains the benchmark to beat, and by its own maker's estimate, Grok's catch-up is a 2027 story.
