Elon Musk's xAI Unveils Grok 3
Grok 3 has been launched by Elon Musk's xAI.
TroibNews 
On Monday, Elon Musk's AI company, xAI, introduced its latest model, Grok 3, bringing new features to both its iOS and web applications. Grok, designed to compete with OpenAI's GPT-4 and Google's Gemini, offers capabilities in image analysis and answering queries.
The release of Grok 3 was pushed back from its anticipated 2024 launch following several months of development. The training for this advanced model took place at xAI's extensive data center in Memphis, which is equipped with 200,000 GPUs. Musk claims that Grok 3 employs "10x" more computing power than Grok 2, benefiting from a broader dataset that includes court case documents. Musk describes Grok 3 as a significantly enhanced, "maximally truth-seeking AI," even when the truth challenges political correctness.
Grok 3 is available in multiple versions, including a faster and smaller variant, Grok 3 mini, which trades off some accuracy for quicker response times. While not all versions have been released yet, the rollout commenced on Monday. According to xAI, Grok 3 surpasses GPT-4o in benchmark tests, such as AIME, which evaluates math problem-solving abilities, and GPQA, which poses PhD-level science questions to AI. Grok 3 also performed admirably in the crowdsourced Chatbot Arena.
Two specialized versions, Grok 3 Reasoning and Grok 3 mini Reasoning, are tailored for problem-solving by thoroughly fact-checking responses to mitigate common errors. These reasoning-focused models are particularly effective for complex math, science, and programming questions, reportedly outperforming even OpenAI's best o3-mini version across various benchmarks, including AIME 2025.
The Grok app is set to introduce a "voice mode" that will feature synthesized speech, followed by the rollout of Grok 3 models in xAI's enterprise API and the DeepSearch functionality. Additionally, Musk intends to open-source Grok 2 in the coming months.
When Grok was initially announced, Musk positioned it as an unfiltered, anti-"woke" AI, prepared to tackle controversial questions that others hesitate to address. The Grok models have delivered on this promise, providing responses that other AIs, like ChatGPT, may avoid, including using bold or vulgar language when requested.
Olivia Brown for TROIB News
The release of Grok 3 was pushed back from its anticipated 2024 launch following several months of development. The training for this advanced model took place at xAI's extensive data center in Memphis, which is equipped with 200,000 GPUs. Musk claims that Grok 3 employs "10x" more computing power than Grok 2, benefiting from a broader dataset that includes court case documents. Musk describes Grok 3 as a significantly enhanced, "maximally truth-seeking AI," even when the truth challenges political correctness.
Grok 3 is available in multiple versions, including a faster and smaller variant, Grok 3 mini, which trades off some accuracy for quicker response times. While not all versions have been released yet, the rollout commenced on Monday. According to xAI, Grok 3 surpasses GPT-4o in benchmark tests, such as AIME, which evaluates math problem-solving abilities, and GPQA, which poses PhD-level science questions to AI. Grok 3 also performed admirably in the crowdsourced Chatbot Arena.
Two specialized versions, Grok 3 Reasoning and Grok 3 mini Reasoning, are tailored for problem-solving by thoroughly fact-checking responses to mitigate common errors. These reasoning-focused models are particularly effective for complex math, science, and programming questions, reportedly outperforming even OpenAI's best o3-mini version across various benchmarks, including AIME 2025.
The Grok app is set to introduce a "voice mode" that will feature synthesized speech, followed by the rollout of Grok 3 models in xAI's enterprise API and the DeepSearch functionality. Additionally, Musk intends to open-source Grok 2 in the coming months.
When Grok was initially announced, Musk positioned it as an unfiltered, anti-"woke" AI, prepared to tackle controversial questions that others hesitate to address. The Grok models have delivered on this promise, providing responses that other AIs, like ChatGPT, may avoid, including using bold or vulgar language when requested.
Olivia Brown for TROIB News
