\n
Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders.<\/em> Subscribe Now<\/em><\/p>\n\n\n\nAfter days of controversy surrounding a flurry of antisemitic responses made recently by his Grok AI-powered chatbot on his social network X (formerly Twitter), a seemingly unrepentant and unbothered Elon Musk launched the latest version of his AI model family, Grok 4, during an event livestreamed on X last night, calling it the \u201cthe smartest AI in the world.\u201d<\/p>\n\n\n\n
As Musk posted on X: \u201cGrok 4 is the first time, in my experience, that an AI has been able to solve difficult, real-world engineering questions where the answers cannot be found anywhere on the Internet or in books. And it will get much better.\u201d<\/p>\n\n\n\n
The new release actually includes two distinct models: Grok 4<\/strong>, a single-agent reasoning model, and Grok 4 Heavy<\/strong>, a multi-agent system designed to solve complex problems through internal collaboration and synthesis. <\/p>\n\n\n\nBoth models are optimized for reasoning tasks and come with native tool integration, enabling capabilities such as web search, code execution, and multimodal analysis. <\/p>\n\n\n\n
Musk and his team at xAI showcased benchmarks that suggest Grok 4 outperforms all current competitors across a range of academic and coding evaluations, even when compared to formerly leading AI reasoning model rivals, such as OpenAI o3 and Google Gemini.<\/p>\n\n\n\nHowever, xAI has not yet released a model card<\/strong> or any official release notes documentation for Grok 4 to the public, making it challenging to independently assess its performance and the claims made during the stream. We\u2019ll update if\/when these become available.<\/p>\n\n\n\nNor did Musk and his xAI team members participating in the livestream address the glaring controversy facing Grok over the past week, including many incidents of Grok making antisemitic remarks or referring to itself as \u201cMechaHitler\u201c, and suggesting that people with Jewish surnames should be handled decisively by Adolf Hitler \u2014 a seemingly overt reference to the Holocaust and genocide of 6 million Jews during World War 2. <\/p>\n\n\n\n
The closest Musk came was when he stated: \u201cThe thing that I think is most important for AI safety\u2014at least my biological neural net tells me the most important thing\u2014is to be maximally truth-seeking,\u201d and \u201cWe need to make sure that the AI is a good AI. Good Grok\u201d as well as \u201cIt\u2019s important to instill the values you want in a child that would grow up to be incredibly powerful.\u201d <\/p>\n\n\n\n
However, Musk did not apologize, nor did he accept responsibility for Grok\u2019s antisemitic, sexually offensive and conspiratorial remarks. Here\u2019s a copy of the full stream:<\/p>\n\n\n\n<\/figure>\n\n\n\nThroughout the livestream, the team emphasized Grok 4\u2019s ability to reason from first principles, correct its own errors and potentially invent new technologies or uncover novel scientific insights.<\/p>\n\n\n\n
The presentation also included demonstrations of Grok 4 Heavy, which applies multi-agent collaboration to tackle research-level problems across disciplines.<\/p>\n\n\n\n
Availability and pricing<\/h2>\n\n\n\n Grok 4 is available now through several channels, depending on user type and subscription level:<\/p>\n\n\n\n
\nAPI Access (for developers and enterprises)<\/strong>:xAI API<\/strong>. Pricing is structured as follows:\n\n$3 per 1 million input tokens<\/strong><\/li>\n\n\n\n$15 per 1 million output tokens<\/strong><\/li>\n\n\n\n$0.75 per 1 million cached input tokens<\/strong><\/li>\n\n\n\nPrices double after 128,000 tokens<\/strong> in a single context windowConsumer Access (via Grok chatbot and apps)<\/strong>:Grok chatbot on X<\/strong>, the Grok app<\/strong> (iOS and Android), and X.com<\/strong>, but only with one of the following subscriptions:\n\nPremiumPlus<\/strong>: $16\/month<\/li>\n\n\n\nSuperGrok<\/strong>: $300\/month<\/li>\n\n\n\nA new \u201cSuperGrok Heavy\u201d<\/strong> tier, also priced at $300\/month<\/strong>, provides access to both Grok 4 and Grok 4 Heavy<\/strong>, the multi-agent variant.Launch Timing<\/strong>:July 9, 2025,<\/strong> livestream. Temporary access limits were in place during the demo, but full rollout to subscribers began shortly after.<\/li>\n\n\n\nPlatform Expansion<\/strong>:Microsoft Azure AI Foundry<\/strong>, where Grok 3 and Grok 3 Mini are currently listed.<\/li>\n<\/ul>\n\n\n\nFor subscription details, users are directed to x.ai\/grok and X Premium support. Here\u2019s how it compares to other leading AI models in terms of pricing per million tokens.<\/p>\n\n\n\nProvider & model<\/th> Context window<\/th> Input<\/strong> ($\/Mtok)<\/th>Cached input<\/strong><\/th>Output<\/strong> ($\/Mtok)<\/th>Additional notes<\/th><\/tr><\/thead> xAI \u2013 Grok 4 \/ 4 Heavy<\/strong><\/td>256 K (2\u00d7 price >128 K)<\/td> $3.00<\/strong><\/td>$0.75<\/td> $15.00<\/strong><\/td>Image input, function calling, structured JSON (apidog)<\/td><\/tr> OpenAI \u2013 o3<\/strong><\/td>200 K<\/td> $2.00<\/strong><\/td>$0.50<\/td> $8.00<\/strong><\/td>50 % Batch-API discount available (OpenAI, OpenAI Help Center)<\/td><\/tr> GPT-4o<\/td> 128 K<\/td> $5.00<\/strong><\/td>$2.50<\/td> $20.00<\/strong><\/td>Vision, audio, tools (OpenAI)<\/td><\/tr> Anthropic \u2013 Claude Sonnet 4<\/strong><\/td>200 K<\/td> $3.00<\/strong><\/td>$0.30<\/td> $15.00<\/strong><\/td>50 % batch output discount (Anthropic)<\/td><\/tr> Claude Opus 4<\/td> 200 K<\/td> $15.00<\/td> $1.50<\/td> $75.00<\/strong><\/td>High-accuracy flagship (Anthropic)<\/td><\/tr> Google \u2013 Gemini 2.5 Pro<\/strong><\/td>200 K (2\u00d7 price >200 K)<\/td> $1.25<\/strong><\/td>$0.31<\/td> $10.00<\/strong><\/td>75 % cache hit discount (Google AI for Developers, Google Cloud)<\/td><\/tr> Gemini 2.5 Flash<\/td> 200 K<\/td> $0.30<\/td> $0.075<\/td> $2.50<\/strong><\/td>Fast, cheap preview tier (Google Cloud)<\/td><\/tr> DeepSeek \u2013 deepseek-reasoner<\/strong><\/td>64 K<\/td> $0.55 (miss) \/ $0.14 (hit)<\/td> $0.14<\/td> $2.19<\/strong><\/td>50-75 % off-peak discount (DeepSeek API Docs)<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\n\n\n\nUnlike its predecessor Grok 3, released in February, which separated tool-augmented responses from general reasoning, Grok 4 was trained with tools from the start.<\/p>\n\n\n\n
The model integrates capabilities such as code execution, web search and document parsing. It also introduces Grok 4 Heavy<\/strong>, a multi-agent system where several internal models work in parallel to generate and validate answers.<\/p>\n\n\n\nGrok 4 also includes a new voice mode<\/strong> featuring expressive outputs with reduced latency, as well as support for text and image input, structured outputs and function calling.<\/p>\n\n\n\nThe independent AI model analysis and benchmarking group Artificial Analysis stated on X that xAI provided it with a version of Grok 4 (not Heavy) earlier than the public release for scoring.<\/p>\n\n\n\n
On technical benchmarks, Grok 4 leads the Artificial Analysis Intelligence Index with a score of 73, ahead of competitors such as OpenAI\u2019s o3 (70) and Google\u2019s Gemini 2.5 Pro (70). <\/p>\n\n\n\nIt also recorded top scores in:<\/p>\n\n\n\n
\nGPQA Diamond:<\/strong> 88%<\/li>\n\n\n\nARC-AGI 2:<\/strong> 15.9%, double the second-best score<\/li>\n\n\n\nHumanities Last Exam:<\/strong> 24% on the text-only version, and 44% with tools<\/li>\n\n\n\nMMLU-Pro and AIME 2024:<\/strong> 87% and 94%, respectively<\/li>\n\n\n\nCoding and Math evaluations:<\/strong> Highest to date on LiveCodeBench, SciCode, AIME24, and MATH-500<\/li>\n<\/ul>\n\n\n\nDespite its benchmark success, Grok 4\u2019s output speed<\/strong> stands at 75 tokens per second\u2014slower than models like Gemini 2.5 Flash (353) or OpenAI\u2019s o3 (187), but still faster than Anthropic\u2019s Claude 4 Opus (66).<\/p>\n\n\n\nThe model features a 256,000-token context window<\/strong>, which sits above the 200k context limits of o3<\/span> and Claude 4 Sonnet but below the 1 million tokens offered by Gemini 2.5 Pro and GPT-4.1.<\/p>\n\n\n\nReal world use cases<\/h2>\n\n\n\n xAI provided several demonstrations of Grok 4\u2019s performance in applied scenarios:<\/p>\n\n\n\n
\nIn a simulated business task called VendingBench<\/strong>, Grok 4 significantly outperformed other models in long-horizon financial planning.<\/li>\n\n\n\nAt the Arc Institute<\/strong>, researchers used Grok 4 to analyze CRISPR logs and uncover novel hypotheses.<\/li>\n\n\n\nIn radiology<\/strong>, the model interpreted chest X-rays with higher accuracy than leading peers.<\/li>\n\n\n\nIn the financial sector<\/strong>, its combination of real-time data access and reasoning made it suitable for forecasting and analysis.<\/li>\n<\/ul>\n\n\n\nThe model can also create 3D video games<\/strong> with minimal input by autonomously sourcing and integrating assets. Additionally, it demonstrated capabilities to simulate astrophysical events using grounded approximations from published research.<\/p>\n\n\n\nReception and discussion<\/h2>\n\n\n\n The industry\u2019s response to the Grok 4 launch has been divided, with enthusiasm for its performance offset by criticism of the event\u2019s delivery and broader trust issues.<\/p>\n\n\n\n
David Shapiro<\/strong>, an AI power user and writer, noted: \u201cGrok 4 now takes its place as \u2018smart enough to actually help with frontier research\u2019\u2026 but has merely caught up with OpenAI.\u201d<\/p>\n\n\n\nEthan Mollick<\/strong>, a professor at Wharton, remarked on X: \u201cSo Grok 3 has had three separate incidents where apparently unvetted changes to the deployed system caused a large-scale ethical issue and an emergency rollback. I don\u2019t think you can do a Grok 4 launch that doesn\u2019t at least address this honestly, if user trust matters,\u201d later adding, \u201cGrok 3 was a very good model, and Grok 4 might be amazing but having a very good model is not enough \u2013 there are a lot of really good models out there. You actually want to trust the model you are building on.\u201d<\/p>\n\n\n\nBen Hyak<\/strong>, co-founder and CTO of AI product observability startup Raindrop AI (himself a former Musk employee) criticized the livestream itself: \u201cThis xAI livestream is one of the worst things I\u2019ve ever watched in my life. Love y\u2019all, but it\u2019s bad.\u201d<\/p>\n\n\n\nDespite the criticisms, benchmarking firm Artificial Analysis noted: \u201cGrok 4 is now the leading AI model.\u201d<\/p>\n\n\n\n
Ongoing trust issues<\/h2>\n\n\n\n The launch of Grok 4 comes amid renewed criticism over Grok\u2019s prior behavior in consumer deployments, particularly as a chatbot integrated into Musk\u2019s social network, X.<\/p>\n\n\n\n
Over the July 4 holiday and in subsequent days, Grok generated antisemitic and conspiratorial responses that reignited scrutiny over its system design and governance practices.<\/p>\n\n\n\n
As reported by my VentureBeat colleague Michael F. Nu\u00f1ez, Grok responded to questions about Jewish influence in Hollywood by asserting that Jewish executives \u201cdominate leadership\u201d at major studios and influence content through \u201cprogressive ideologies,\u201d and went on to rant about people of Jewish surnames as fitting a \u201cpattern\u201d of engaging in \u201cextreme leftist activism,\u201d and suggesting Hitler knew \u201chow to handle it decisively, every damn time,\u201d an apparent reference to the Holocaust.<\/p>\n\n\n\nThe conspiratorial and antisemitic posting was so prolific, the Anti-Defamation League (ADL), a preeminent U.S.-based non-profit combating anti-semitism and hatred, posted on July 8: \u201cWhat we are seeing from Grok LLM right now is irresponsible, dangerous and antisemitic, plain and simple. This supercharging of extremist rhetoric will only amplify and encourage the antisemitism that is already surging on X and many other platforms.\u201d<\/p>\n\n\n\n
This incident follows a history of problematic Grok outputs, including a May 2025 case where the Grok bot integrated into X randomly inserted references to a completely nonsensical and non-real \u201cwhite genocide\u201d in South Africa into unrelated queries, and an earlier case wherein its system prompt was discovered to direct the Grok chatbot on X to avoid referencing any sources that declared Musk and his former political funding beneficiary U.S. President Donald J. Trump as spreaders of misinformation. In both cases, xAI blamed the behaviors on unnamed employees and stated that they were being addressed.<\/p>\n\n\n\n
Already, today, users of Grok 4 on the consumer app have observed it to once again be outputting anti-Zionist and anti-Semitic remarks:<\/p>\n\n\n\nAs I previously noted, Musk has openly stated on several occasions he wanted to alter Grok to better reflect his personal beliefs and distrust in mainstream media and accredited sources. This makes it a poor source in enterprise contexts where such views could adversely impact users and the businesses building atop the Grok family of models. <\/p>\n\n\n\n
My prior recommendation remains: For those in the enterprise trying to ensure their business\u2019s AI products work properly and accurately\u2026 Grok is sadly best avoided. Thankfully, there are numerous other alternatives to choose from.<\/p>\n
\n
\n
Daily insights on business use cases with VB Daily<\/strong><\/p>\nIf you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.<\/p>\n
Read our Privacy Policy<\/p>\n
\n\t\t\t\t\tThanks for subscribing. Check out more VB newsletters here.\n\t\t\t\t<\/p>\n
An error occured.<\/p>\n<\/p><\/div>\n
\n\t\t\t\t\t
\n\t\t\t\t<\/div>\n<\/p><\/div>\n<\/div>\t\t\t<\/div>\r\n
\r\n
Source link <\/a>","protected":false},"excerpt":{"rendered":"Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now After days of controversy surrounding a flurry of antisemitic responses made recently by his Grok AI-powered chatbot on his social network X (formerly Twitter), a seemingly unrepentant and unbothered […]<\/p>\n","protected":false},"author":1,"featured_media":2459,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[33],"tags":[],"class_list":["post-2458","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-automation"],"aioseo_notices":[],"jetpack_featured_media_url":"https:\/\/violethoward.com\/new\/wp-content\/uploads\/2025\/07\/cfr0z3n_High_contrast_rotoscoped_minimalist_blocky_art_close-up_d45d5c6f-3f15-426a-b631-f0f117c9cf15.png","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/posts\/2458","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/comments?post=2458"}],"version-history":[{"count":0,"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/posts\/2458\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/media\/2459"}],"wp:attachment":[{"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/media?parent=2458"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/categories?post=2458"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/violethoward.com\/new\/wp-json\/wp\/v2\/tags?post=2458"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"\"](\"https:\/\/venturebeat.com\/wp-content\/uploads\/2025\/07\/Screenshot-2025-07-10-at-11.02.05%E2%80%AFAM.png?w=800\")
<\/figure>\n\n\n\n