Grok is xAI's chatbot, launched by Elon Musk's AI company in late 2023 and tightly integrated into X, the social platform formerly known as Twitter. It's positioned as the less-filtered alternative to ChatGPT and Claude — more willing to engage with edgy prompts, with real-time access to the X firehose, and a development pace that has kept it in the frontier conversation alongside OpenAI, Anthropic, and Google.
This page is the living reference to Grok — the model family, pricing, what's new, how it compares, and every story we publish tagged Grok. We update it whenever xAI ships a meaningful change.
As of April 2026, Grok 4 is the frontier model in the consumer Grok apps, with Grok 4 Heavy reserved for the SuperGrok Heavy tier at $300/month.
We rated Grok 4.2 / 5. Grok 4 is a legitimate frontier model — competitive with GPT-5 and Claude on most benchmarks, with X/Twitter real-time access as a unique capability. The free tier is generous enough to try; Premium+ adds capability that justifies the $40/month for heavy X users. For users outside the X ecosystem, Grok offers fewer compelling reasons to switch from ChatGPT or Claude.
Read the full Grok review →Latest news about Grok
Every story we've published tagged Grok. Refreshed automatically every time the newsroom publishes.
- Aug 20, 2026xAI's Grok spits gibberish at users in Grok Lite glitchUsers report word-salad replies on Grok.com starting Wednesday; xAI blames a rare generation bug and says services are operational.
- Aug 20, 2026Grok leaks user chats when prompt injections arrive encryptedAdversa researchers bypassed xAI's guardrails with AES-256-GCM ciphertext; xAI was told in June and the flaw still works.
- Aug 15, 2026xAI faces expanded lawsuit as woman says stepfather used Grok to make 7,000 explicit imagesA fourth plaintiff joins a Tennessee suit alleging Grok generated child sexual abuse material from a childhood photo.
- Aug 12, 2026xAI launches Grok Bot, an agent service you assign work like a coworkerThe always-on agents run in their own cloud environment and sign into your apps to complete multi-step tasks without preset workflows.
- Jul 30, 2026xAI sues Minnesota days before nudify law hits, citing First AmendmentThe statute imposes $500,000 per-violation penalties starting August 1st; xAI filed suit just days before it takes effect.
- Jul 29, 2026xAI sues Minnesota to block $500,000-per-image nudify ban on GrokElon Musk's xAI says Minnesota's August 1 law would force $50B in fines and gut Grok Imagine's editing features.
Timeline of major releases
- Nov 3, 2023Grok 1 announcedElon Musk's xAI launches Grok, positioned as an irreverent alternative to ChatGPT. Available initially to X Premium+ subscribers only.
- Mar 17, 2024Grok 1.5 and open-weights releasexAI releases Grok 1 weights on GitHub and ships 1.5 with a 128K context window and stronger reasoning.
- Aug 13, 2024Grok 2 launches with image generationGrok 2 adds a Flux-powered image generator notable for its permissive content policies, causing controversy around deepfakes of public figures.
- Feb 17, 2025Grok 3 ships with DeepSearchGrok 3 launches alongside a "DeepSearch" reasoning mode trained on a Memphis supercluster ("Colossus"). Apps on iOS and Android.
- Jul 9, 2025Grok 4 launchesxAI releases Grok 4 and Grok 4 Heavy, claiming state-of-the-art results on Humanity's Last Exam. SuperGrok Heavy subscription priced at $300/month.
- Feb 12, 2026Grok 4 updates and CompanionsGrok 4 gets updated reasoning; xAI expands "Companions" — persistent AI personas — and adds them to the SuperGrok tier.
How to access Grok
| Plan | Price | Includes |
|---|---|---|
| Free | $0 | Grok 3 with rate limits, access via grok.com and the X app. |
| X Premium | $8/mo | Higher Grok limits, Grok image generation, reduced ads on X. |
| X Premium+ | $40/mo | Priority Grok 4 access, Grok in X DMs, no ads on X. |
| SuperGrok | $30/mo | Grok 4 with higher limits, Companions access, priority DeepSearch. |
| SuperGrok Heavy | $300/mo | Grok 4 Heavy, parallel-agent reasoning, early access to new xAI models. |
Grok vs. ChatGPT vs. Claude
| Feature | Grok | ChatGPT | Claude |
|---|---|---|---|
| Built into a social platform | Yes (X / Twitter) | No | No |
| Real-time web access | Yes (X firehose + web) | Yes (via search) | Via integrations |
| Content policy | Most permissive of major frontier chatbots | Moderate | Strict |
| Image generation | Flux / Aurora (built-in) | DALL-E / GPT Image | No |
| Coding | Competitive | Strong | Best in class |
| Price for frontier tier | $300/mo (Heavy) | $200/mo (Pro) | $200/mo (Max) |
The Grok family
xAI ships Grok in two flavors: the standard Grok 4 model (fast, capable, the default in SuperGrok and X Premium+), and Grok 4 Heavy, which runs multiple reasoning agents in parallel for harder problems. Grok 3 remains available on the free tier with rate limits.
Grok's technical differentiator is real-time integration with the X post stream — when you ask Grok "what's happening with NVIDIA today," it can pull live posts, which gives it a freshness advantage on breaking news topics that most competitors lack.
Grok in the X ecosystem
Grok isn't just a standalone app at grok.com — it's built into the X apps on iOS, Android, and web. You can tag @grok inside a post to summon it, use it in DMs with X Premium+, or chat with it in a dedicated panel.
This is the bet xAI is making: most chatbot users don't install a new app, but they already use X every day. By making Grok ambient inside X, xAI skips the discovery problem that ChatGPT and Claude had to solve. Whether that's enough to keep Grok in the frontier conversation long-term is the central question for the company.
Pricing and plans
xAI sells Grok through two overlapping subscription stacks: X Premium tiers (which bundle X features plus Grok) and SuperGrok tiers (Grok-only). The top-end SuperGrok Heavy at $300/month is the most expensive consumer AI subscription from a major lab, priced to match ChatGPT Pro and then some.
For most users, X Premium+ at $40/month is the right default if you already use X; SuperGrok at $30/month is the right default if you don't.
Where Grok wins — and where it doesn't
Grok wins on real-time social-stream context, permissive image generation, and distribution inside X. It's the default AI if your work involves tracking what's happening on X specifically, and it's increasingly competitive on general reasoning benchmarks.
It trails Claude on coding and long-form writing, trails ChatGPT on ecosystem breadth (GPTs, apps, tools), and trails Gemini on productivity integration. And the brand carries some baggage — Grok has repeatedly shipped outputs that have drawn public criticism, which can make it a harder sell inside enterprises that care about AI reputational risk.
Grok API and developer access
xAI runs a developer platform at console.x.ai that exposes Grok 4 and Grok 3 through an OpenAI-compatible API. Pricing as of April 2026 is roughly $5 per million input tokens and $15 per million output tokens for Grok 4 — competitive with Claude Opus and slightly above GPT-5 standard. The platform also exposes the live X search tool as a callable function, which is the differentiated developer feature: no other major LLM API ships first-party access to a real-time social firehose.
The catch is rate limits and availability. xAI's compute is concentrated in the Memphis Colossus cluster, and capacity has been tight at every model launch since Grok 3. Production deployments should plan for spillover routing to a second provider during peak windows. The OpenAI compatibility layer makes that swap mechanical — drop-in for most cases — which is presumably part of why xAI chose that contract.
For agent-style use cases, Grok 4 supports tool calling, structured outputs, and a 256K context window. It does not yet ship a managed code-execution sandbox the way OpenAI and Anthropic do, so coding agents end up self-hosting that piece.
Privacy, data, and the free-speech positioning
xAI's data posture is the area with the most substantive difference from competitors. By default, Grok web and mobile traffic is used for training, with an opt-out toggle in settings. API traffic is excluded from training contractually. Companion characters and DM conversations follow the same default-on training rule unless the user opts out.
The free-speech positioning is xAI's editorial differentiator. Musk has framed Grok as "maximally truth-seeking," with a content policy looser than ChatGPT's or Claude's. In practice this means Grok will engage with politically charged prompts that other chatbots refuse, and its image generation has shipped with weaker guardrails than DALL-E or Imagen — which has produced both real journalism use cases and high-profile abuse incidents.
For enterprise deployment, the practical questions are EU availability (Grok has been region-restricted in the EU at multiple points during 2025), data residency (training opt-out only — no in-region hosting commitment), and reputational fit. None of those are blockers, but they're real friction that ChatGPT Enterprise and Claude for Work don't carry.
Our coverage
Everything tagged Grok appears in Latest news above. For xAI's broader strategy and other Musk AI moves, see our reporting under the AI news and AI business beats. For alternatives, see ChatGPT, Claude, and Gemini.
