All Grok models available in 2025: full list for web, app, and API including Grok 4, 3 Mini, and image generation
- Graziano Stefanelli
- 2 days ago
- 4 min read
Grok now powers AI conversations across X (Twitter), web, and developer APIs with high-context and search-augmented models.
As of August 2025, Grok by xAI has solidified its multi-tier model lineup for consumers and developers, delivering models with expansive context windows, search-grounded outputs, and multimodal capabilities. Users accessing Grok through the X platform or its standalone app are routed through Grok 4, while developers integrating via API can choose from multiple models including Grok 3, 3 Mini, and Grok 2 Image.
Here we share a full breakdown of all Grok models currently available via web, mobile, and API, with a focus on version IDs, capabilities, and context.
Grok on the web and mobile apps uses Grok 4 and Heavy variant for premium tiers.
Grok is embedded inside X (formerly Twitter) and also available as a standalone product via grok.com and native apps for iOS and Android. The default AI experience for most users is powered by Grok 4, with an optional upgrade to Grok 4 Heavy for SuperGrok subscribers.
● Grok 4
Announced on July 9, 2025, Grok 4 is the standard model behind all X Premium+ and SuperGrok accounts, delivering responses with up to 256,000 tokens of context. It supports enhanced reasoning, humor-infused tone, and structured outputs, especially for queries about real-time events, politics, or tech culture.
● Grok 4 Heavy
Available exclusively through the SuperGrok Heavy tier, this model offers higher fidelity responses, improved memory consistency, and faster real-time search integration. It is accessible only through the app/web UI, and is not yet exposed via API. The distinction is handled at the account tier level, not by manual model selection.
xAI provides full Grok API access with multiple models and capabilities.
The Grok API, documented on x.ai, exposes several LLMs with clear model IDs, context lengths, and purpose-built functionalities. Developers can integrate these models into applications ranging from conversational agents to content generators and search-enhanced knowledge bots.
Model Name | Model ID | Token Context | Purpose |
Grok 4 | grok-4-0709 | 256,000 | Flagship model for reasoning, conversation, synthesis |
Grok 3 | grok-3 | 131,072 | Legacy model with solid general capabilities |
Grok 3 Mini | grok-3-mini | 131,072 | Optimized for cost-efficiency and fast completions |
Grok 2 Image | grok-2-image-1212 | N/A | Dedicated for image generation from text prompts |
Grok 4 is the most powerful model available via API, with extensive reasoning capabilities and a long context window ideal for structured generation, Q&A over documents, or prompt chaining. All models support standard Chat Completions-style API calls with customizable temperature and response control.
Vision and search capabilities expand Grok’s reach across modalities.
● Image generation with Grok 2 Image
The grok-2-image-1212 endpoint powers text-to-image generation, producing stylized images on demand. While not the most recent model in the LLM family, it allows developers to extend Grok into visual storytelling, product mockups, or illustrated answers.
● Real-time Live Search API
Grok also offers a Live Search integration via API, allowing access to indexed sources from X, the open web, and verified databases. Developers are charged $25 per 1,000 sources retrieved (or $0.025 per source), and the data is streamed as context to the model in real time.
This makes Grok a strong choice for search-augmented generation (RAG) and news-based use cases.
Knowledge cutoff and search hybridization affect output behavior.
Grok 4 and 3 models have an internal knowledge cutoff of November 2024. However, when paired with Live Search, they can deliver fully up-to-date answers, referencing new articles, posts, or research papers—especially across trending topics on X. Search results are passed into the prompt chain to provide citations or inline data.
Platform access and selection rules vary by interface.
Platform | Model(s) Used | How It’s Selected |
Grok on X / Web / App | Grok 4 or Grok 4 Heavy | Routed automatically by subscription tier |
xAI API (developer access) | Grok 4, 3, 3 Mini, 2 Image | Selected manually via model ID |
Image tasks via API | Grok 2 Image | Dedicated endpoint |
Real-time search | Grok 3 / 4 + Live Search API | Activated via parameters |
There is currently no public model selector inside the Grok web interface. Only API developers can manually invoke a specific model by passing the correct model ID in their request payload.
Choosing the right Grok model depends on your use case and deployment method.
For consumer chat experiences with humor, web search, and long context → use Grok 4
For premium interactive sessions (long memory, higher priority) → upgrade to Grok 4 Heavy
For budget-conscious apps or fast agents → deploy Grok 3 Mini
For creative visual content → use Grok 2 Image via API
For news tracking, real-time citations, and RAG → combine Grok 4 + Live Search
Grok now stands as a unique offering in the AI chatbot space—balancing personality, performance, and programmability. It is particularly favored by users seeking culturally aware responses, real-time web grounding, or open access to high-context LLMs with minimal friction.
____________
FOLLOW US FOR MORE.
DATA STUDIOS