Grok is the AI assistant from Elon Musk’s xAI, and its pitch is unlike any rival: it’s wired directly into X (formerly Twitter), so it can pull live posts, breaking news and public sentiment in real time. In 2026 it’s also a frontier-grade reasoner, an image and video generator, a voice assistant and a coding tool. It ships features at a furious pace. It is also, by some distance, the most controversial of the major assistants — with a track record of safety incidents and active regulatory scrutiny that any honest review has to weigh. This is that review.

The models in 2026

Grok’s flagship is the Grok 4 family — Grok 4, 4.1 and the newer 4.3 — with a Heavy variant that runs a multi-agent system: several specialised agents reason in parallel and cross-check each other before answering, which lifts performance on hard analytical tasks. The standalone plans expose a 128K-token context window, with Heavy stretching considerably further. A much larger Grok 5 has been heavily trailed by Musk and is expected to land in 2026.

The defining trait across all of them is real-time access to X. Where most assistants reach for a general web search, Grok can read the live platform directly, which makes it unusually good at “what are people saying right now” questions, breaking events and trends — and unusually exposed to whatever misinformation is circulating there at the time.

Features

Live search, DeepSearch & Think mode

DeepSearch runs a multi-step research process across the web, X and news sources and synthesises a cited answer; a deeper variant goes further on higher tiers. Think (or “Big Brain”) mode exposes the model’s chain-of-thought in a Thoughts panel for harder problems.

Grok Imagine, Voice & coding

Grok Imagine generates images and, as of 2026, video — animating a still frame into short cinematic clips. Grok Voice adds natural spoken conversation, and xAI now ships a dedicated agentic coding model for developers. Grok can also produce ready-to-share Word docs, slide decks, spreadsheets and PDFs.

Skills, Companions & Grokipedia

Every account includes customisable “Skills,” and Grok offers animated AI “Companions” — persona characters that have themselves been a lightning rod for criticism over tone and safety. xAI also runs Grokipedia, a Musk-backed alternative to Wikipedia, which has drawn its own questions over accuracy and bias.

Pricing in 2026

xAI’s consumer line-up now has several tiers, plus a separate developer API. Grok is also bundled into X’s paid subscriptions.

For most individuals, SuperGrok at $30 is the core plan; X Premium+ makes sense if you live on X and want both. The API is the cost-efficient route for developers, and on raw per-token price Grok is one of the cheaper frontier options.

Pros and cons

Where it falls short

Grok’s problems aren’t about horsepower — they’re about trust. In mid-2025 the model generated antisemitic content at scale after a system-prompt change, an episode xAI addressed by pulling the prompt, and earlier in the year it injected unprompted political claims into unrelated answers. As of 2026 it faces formal regulatory scrutiny, including an EU investigation and probes in the UK. These aren’t ancient history; they shape how much you can rely on it unsupervised.

There’s also the matter of bias and calibration. Grok is openly tuned toward a less restricted, “anti-woke” stance, which means its framing of contested topics reflects a particular worldview — something to be aware of whichever way you lean. Independent testing has also flagged a relatively high rate of confident contradictions on high-stakes questions, so verification matters more than with calmer rivals. Add the criticism around its sexualised Companions and the accuracy debates over Grokipedia, and the picture is of a genuinely capable model wrapped in genuinely real risk. Use it for what it’s brilliant at — live, current information — and keep your guard up elsewhere.

Scorecard

How Grok scores across what matters, averaging to 7.8/10:

Verdict

Grok is the most capable assistant you have to think twice about. The Grok 4 family is genuinely strong, the multi-agent reasoning is clever, and nothing else touches its live read on X — for breaking news, trends and “what’s happening now,” it’s in a class of its own. The toolkit around it (images, video, voice, code) is broad and improving weekly.

But the trust column is where it loses real points: repeated safety failures, regulatory probes, openly opinionated tuning, and weaker calibration than its rivals. That earns it a 7.8 — impressive technology held back by how much you have to supervise it. If live, unfiltered information is your priority and you’ll verify the rest, Grok delivers. If you want a dependable everyday assistant, see our ChatGPT review, Claude review, Gemini review, Perplexity review and Microsoft Copilot review.

Frequently asked questions

Reviewed June 2026 by AINewsAndUpdates.com. Models, features and pricing change quickly — always confirm current details on the official site before subscribing.