February 23, 2025 – Elon Musk’s AI company, xAI, has launched its latest flagship AI model, Grok 3, unveiling new capabilities for its Grok iOS and web applications.
Grok 3, positioned as a competitor to OpenAI’s GPT-4o and Google’s Gemini, offers enhanced reasoning capabilities and image analysis, further integrating with Musk’s social network, X.
Grok 3: enhanced AI capabilities
Grok 3 has been in development for months and was initially slated for release in 2024 but experienced delays. The model was trained using an enormous data center in Memphis containing approximately 200,000 GPUs. According to Musk, the new model was developed with “10x” more computing power than its predecessor, Grok 2, leveraging a more extensive training dataset, including legal filings.
“Grok 3 is an order of magnitude more capable than Grok 2,” Musk stated during a live-streamed presentation. “[It’s a] maximally truth-seeking AI, even if that truth is sometimes at odds with what is politically correct.”
Grok 3 model variants and features
The Grok 3 family includes multiple versions:
Grok 3 mini: Offers faster response times at the cost of some accuracy.
Grok 3 Reasoning & Grok 3 mini Reasoning: Designed to carefully “think through” problems, similar to OpenAI’s o3-mini and DeepSeek’s R1. These models aim to fact-check themselves before responding, improving accuracy.
Grok 3 is currently rolling out in phases, with some features still in beta. The new models reportedly outperform GPT-4o in AI benchmarks such as AIME (mathematics performance) and GPQA (graduate-level science questions). Additionally, xAI claims that Grok 3 Reasoning surpasses OpenAI’s o3-mini-high in mathematics assessments, including AIME 2025.
New capabilities: DeepSearch and Reasoning enhancements
The reasoning models power DeepSearch, xAI’s AI-powered research tool, which scans the internet and X to generate detailed abstracts in response to user queries. Users can activate Grok 3’s enhanced reasoning through:
“Think” mode: For complex questions requiring deeper analysis.
“Big Brain” mode: A high-compute setting designed for advanced problem-solving.
To prevent unauthorized knowledge extraction, xAI has partially obscured the reasoning models’ thought processes in the Grok app. This measure follows recent controversies, such as DeepSeek’s alleged distillation of OpenAI models.
Subscription tiers and upcoming features
Grok 3 will first be available to X Premium+ subscribers ($22/month), while advanced features will require a new SuperGrok plan, reportedly priced at $30/month or $300/year. This premium tier grants additional reasoning queries, DeepSearch access, and unlimited image generation.
Upcoming enhancements include:
Voice Mode (within a week): A synthesized voice for Grok models.
Enterprise API Access (within weeks): xAI will offer Grok 3 and DeepSearch via API for business integration.
Future Open-Source plans and political neutrality pledge
Musk announced that xAI intends to open-source Grok 2 in the coming months, following a pattern where previous models are made public once new versions become stable.
“When Grok 3 is mature and stable, which is probably within a few months, then we’ll open-source Grok 2,” Musk confirmed.
Grok was initially marketed as an “edgy” and “unfiltered” AI model that could answer controversial questions other models avoided. However, past versions hedged on political topics, with one study suggesting a left-leaning bias on issues such as transgender rights and diversity programs. Musk has attributed this to training data limitations and has pledged to shift Grok toward greater political neutrality. Whether xAI has achieved this goal remains unclear, but the company continues to push boundaries in AI development.
As Grok 3 rolls out, its real-world performance and reception will determine its place in the competitive AI landscape.
Ship traffic through the Bab el-Mandeb Strait fell to its lowest level in months on Sunday after Yemen's Houthi movement attacked Saudi oil facilities...
The chief executive of a Turkish defence company has said the United States has approved the export of GE F110 engines for Türkiye’s KAAN fighter j...
The number of confirmed Ebola cases in the Democratic Republic of Congo has risen to 3,200, including 1,405 deaths, according to government data relea...
Russian authorities said 11 people, including four children, had been killed by Ukraine in an overnight attack on a holiday camp, accusing Kyiv on Saturday of deliberately targeting civilians.
Iran will halt its own attacks as long as the United States maintains its pause on strikes, but Tehran remains sceptical of U.S. intentions, a senior Iranian source told on Sunday.
Ship traffic through the Bab el-Mandeb Strait fell to its lowest level in months on Sunday after Yemen's Houthi movement attacked Saudi oil facilities along the Red Sea coast, while vessel movements through the Strait of Hormuz also remained subdued despite a pause in U.S.-Iran tensions.
Kazakhstan's President Kassym-Jomart Tokayev called on Saturday for Russia and Ukraine to freeze the conflict and return to peace talks, saying a negotiated settlement remained possible despite the entrenched positions of both sides.
Marvel Studios used its return to San Diego Comic-Con to reveal several major projects, including a new Ghost Rider film starring Ryan Gosling and the long-awaited Black Panther 3.
A New York school district has put plans to introduce a humanoid robot into classrooms on hold after fierce criticism from teachers, Indigenous community members and state education officials over the use of artificial intelligence in schools.
SpaceX successfully launched its Starship megarocket on its 13th flight test Friday, deploying 20 next-generation Starlink V3 satellites and completing a soft splashdown of the spacecraft in the Indian Ocean.
Britain's new Prime Minister Andy Burnham has appointed Kanishka Narayan as Minister for Artificial Intelligence, making it the first time the role has held cabinet-level status. The move signals a stronger government focus on AI as the UK seeks to strengthen its position in the sector.
Scientists have identified a planetary system in the Milky Way unlike anything seen before, leaving astronomers debating how to classify one of its unusual worlds. The discovery challenges long-held definitions of planets, moons and stars.
A U.S. federal judge has approved Anthropic's $1.5 billion settlement with authors and publishers over the company's acquisition and use of pirated books to train its Claude artificial intelligence chatbot, marking the first major resolution in the growing wave of AI copyright litigation.
You can download the AnewZ application from Play Store and the App Store.
What is your opinion on this topic?
Leave the first comment