Elon Musk’s AI firm, xAI, has unveiled Grok 4.1, the latest model of its AI mannequin. The replace introduces a number of enhancements, together with a big discount in hallucinations—false or deceptive outputs—by practically thrice in comparison with earlier variations.
Musk introduced the discharge on his platform X, highlighting noticeable boosts in pace and general efficiency. “Grok 4.1 simply launched. It’s best to discover a big improve in pace and high quality,” he wrote.
Add Zee Information as a Most well-liked Supply
In accordance with the corporate, Grok 4.1 has been refined to take care of robust intelligence and reliability whereas providing enhanced inventive, emotional, and collaborative capabilities.
Grok 4.1 Is 3x Much less More likely to Hallucinate
xAI emphasised that one of many largest enhancements in Grok 4.1 is its diminished tendency to current false info as factual. Tackling hallucinations was a significant focus throughout the mannequin’s post-training course of, particularly for information-based queries.
To measure this, xAI used real-world person prompts and the FActScore benchmark, which comprises 500 biography-related questions.
The sooner Grok 4 Quick mannequin confirmed a hallucination fee of 12%,
whereas Grok 4.1 introduced it all the way down to 4%, a threefold lower.
FactScore outcomes confirmed an analogous enchancment: Grok 4 Quick had a rating of 9.89%, whereas Grok 4.1 achieved 2.97%.
Grok 4.1 Tops Main AI Benchmarks
xAI additionally examined the mannequin on LMArena, a trusted analysis platform for big language fashions. Within the Textual content Area, Grok 4.1 (quasarflux mode) earned the very best general Elo rating of 1483, surpassing all non-xAI fashions by a margin of 31 factors. Even in its less complicated tensor mode, Grok 4.1 secured the second-highest rating, outperforming the full-reasoning modes of many opponents.
The rollout adopted a silent launch part between November 1 and 14, 2025, throughout which the mannequin was steadily launched to extra customers throughout all platforms. All through this era, xAI performed blind pairwise comparisons of Grok 4.1 towards its earlier variations utilizing dwell person site visitors.
In these evaluations, Grok 4.1 achieved a 64.78% win fee, exhibiting a powerful choice amongst customers for the up to date mannequin.
How one can Entry Grok 4.1
Grok 4.1 is now out there to all customers on grok.com, X, and the cellular apps for each iOS and Android. To make use of it, merely choose “Grok 4.1” from the mode picker or allow Auto mode within the app.
With the launch of Grok 4.1, xAI indicators its dedication to constructing sooner, smarter, and extra dependable AI techniques. Because the mannequin rolls out to customers worldwide, its improved accuracy, diminished hallucinations, and powerful benchmark efficiency spotlight a significant leap ahead for the Grok platform. As AI competitors intensifies, Grok 4.1 positions itself as a strong contender—pushing the boundaries of what next-generation AI fashions can ship.