> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zerotwo.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Models Overview
> ZeroTwo provides access to 95+ AI models from 19 providers — text, image, video, and audio. Switch models freely in any chat.
ZeroTwo gives you access to **95+ AI models from 19 providers** in a single interface. No separate accounts, no API keys to manage — pick the model that fits your task and start working. Switch models mid-conversation at any time; the new model receives the full conversation history.
Models are organized into four types: **Text/Chat**, **Image**, **Video**, and **Audio**. Text models power chat conversations; image, video, and audio models are available in chat and in the dedicated Studio pages.
Model access depends on your plan. Standard models are available on all plans. Premium (rate-limited) models require Pro or above. See [Plans and Pricing](/overview/plans-and-pricing) for quota details.
## Model Types
### Text / Chat Models
Text models handle all conversation-based tasks: reasoning, writing, coding, analysis, research, summarization, translation, and more. ZeroTwo hosts text models from 17+ providers.
* **Reasoning models** — step-by-step logical thinking, math, complex coding
* **Writing models** — long-form documents, tone control, nuance
* **Coding models** — code generation, debugging, refactoring, explanation
* **Analysis models** — document understanding, data interpretation, research
### Image Generation Models
Image models are available in the Studio at `/studio/images` and via the Image tool pill in chat. Choose from photorealistic, artistic, multilingual, and specialized generation styles.
### Video Generation Models
Video models are available at `/studio/video`. Generate short clips from text prompts or animate images. Video generation runs as a background job; outputs are saved to your files library.
### Audio Models
Audio models are available at `/studio/audio` and as file attachments in chat. Generate speech, music, and sound effects; automatically transcribe spoken audio via Whisper.
## Provider Overview: Text Models
| Provider | Notable Models | Strengths | Premium? |
| -------------- | ------------------------------------------------------------------- | ------------------------------------------------------------- | ------------------------------------------------------------ |
| **OpenAI** | GPT-5, GPT-4.1, GPT-4o, o3, o4-mini | Coding, instruction-following, reasoning, broad capability | Mostly premium (GPT-5, GPT-4o); standard (GPT-4.1, minis) |
| **Anthropic** | Claude Sonnet 4.6, Claude Opus 4.6, Claude Haiku 4.5 | Writing, analysis, long context (200k tokens), safety-focused | Most are premium (Sonnet, Opus); standard (Haiku) |
| **Google** | Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 3.1 Pro, Gemini Flash Lite | Multimodal, massive context (1M tokens), fast inference | 2.5 Pro and 3.1 Pro are premium; Flash variants are standard |
| **Mistral** | Magistral Latest/Medium, Mistral Large, Mistral Small, Nemo | European privacy focus, multilingual, efficient | Large is premium; Small, Magistral, Nemo are standard |
| **DeepSeek** | DeepSeek Chat, DeepSeek Reasoner, DeepSeek Coder | Coding, math, reasoning, cost-effective | Reasoner is premium; Chat and Coder are standard |
| **Cohere** | Command A, Command R, Command R+, Command R7b | Enterprise search, RAG, document analysis, retrieval | Command A is premium; R variants are standard |
| **xAI (Grok)** | Grok-3, Grok-4, Grok Code Fast | Real-time data access, coding, research, speed | Grok-4 is premium; Grok-3 and Grok Code Fast are standard |
| **Perplexity** | Sonar, Sonar Pro | Web-grounded answers with citations, real-time research | Sonar Pro is premium; Sonar is standard |
| **Qwen** | Qwen Max, Qwen Plus, Qwen Turbo, Qwen Flash, Qwen3 variants | Multilingual, strong Chinese-language support, efficient | Qwen Max is premium; Plus, Turbo, Flash, Qwen3 are standard |
| **Groq** | Various models via Groq inference | Ultra-low latency, speed-first inference | Standard |
| **OpenRouter** | Wide variety via OpenRouter gateway | Access to rare and experimental models | Varies |
| **Kimi K2** | Kimi K2 Thinking, Kimi K2 Turbo, Kimi K2.5 | Long-context document tasks, Chinese-language | Standard |
| **Venice** | Venice Uncensored models | Privacy-focused inference | Standard |
| **TheSys** | TheSys models | Specialized enterprise tasks | Standard |
| **ZAI** | ZAI GLM 4.6 | ZeroTwo-optimized models | Standard |
| **Inception** | Inception models | Specialized research tasks | Standard |
| **ByteDance** | ByteDance GLM 4.7 | Multilingual, content generation | Standard |
## Image Generation Models
ZeroTwo's image studio at `/studio/images` supports multiple generation models with different style characteristics and quality levels.
| Model | Provider | Strengths |
| --------------------------------------- | ----------------- | -------------------------------------------------------- |
| **GPT-Image-1** | OpenAI | Photorealistic outputs, strong instruction-following |
| **GPT-Image-1.5** | OpenAI | Enhanced realism and fine detail over GPT-Image-1 |
| **GPT-Image-mini** | OpenAI | Fast, lighter-weight image generation |
| **Grok Imagine** | xAI | Creative and stylized generations |
| **Grok Imagine Pro** | xAI | Higher quality stylized and photorealistic outputs |
| **Flux Pro v1.1** | Black Forest Labs | High-quality artistic outputs, strong composition |
| **Flux Pro v2** | Black Forest Labs | Latest Flux generation — improved consistency and detail |
| **Imagen 4.0** | Google | Photorealism, excellent text rendering within images |
| **Qwen Image** | Alibaba | Multilingual prompt support, good instruction-following |
| **Qwen Edit** | Alibaba | Image editing and modification from prompts |
| **LustIFY SDXL** | — | SDXL-based generation for artistic styles |
| **Klingai, Creatify, FAL-based models** | Various | Specialized generation options |
You can trigger image generation inline in chat via the Image tool pill in the prompt bar. The dedicated studio at `/studio/images` provides more controls, generation history, and full-resolution outputs with access to all image models.
## Reasoning Models
Several models in ZeroTwo support **extended reasoning** — they work through a problem step-by-step before producing a final answer. When you select a reasoning-capable model, ZeroTwo shows a **thinking level slider** in the prompt bar.
**Reasoning-capable models:**
* **OpenAI o3** — OpenAI's strongest reasoning model
* **OpenAI o4-mini** — faster, more efficient reasoning
* **DeepSeek Reasoner** — strong math and logical reasoning
* **Claude Sonnet 4.6, Claude Opus 4.6** — extended thinking mode available
**Thinking levels:**
| Level | Behavior | Best For |
| ---------- | ----------------------------------------- | -------------------------------------------------- |
| **Low** | Minimal internal reasoning, fast response | Simple questions, quick iteration |
| **Medium** | Balanced reasoning depth and speed | Everyday tasks, moderate complexity |
| **High** | Deep, thorough step-by-step reasoning | Complex math, multi-step coding, detailed analysis |
Higher thinking levels use more tokens and take longer, but produce more accurate and thorough answers for difficult problems. On plans with premium quotas, high-level reasoning responses consume more of your quota.
**When to use reasoning models:**
* Complex math, statistics, and proofs
* Algorithmic coding challenges and debugging multi-file logic
* Structured decision-making and logical analysis
* Research synthesis requiring careful integration of multiple sources
## Premium vs. Standard Models
### Premium (rate-limited) Models
These are the highest-capability, most in-demand models. Sending a message with a premium model counts against your monthly quota on Pro and Pro 2x plans:
* **Claude Opus 4.6 variants, Claude Sonnet 4.6 variants** (Anthropic)
* **GPT-5, GPT-4o** (OpenAI)
* **Gemini 2.5 Pro, Gemini 3.1 Pro** (Google)
* **Grok-4** (xAI)
* **Cohere Command A** (Cohere)
* **Mistral Large** (Mistral)
* **Qwen Max** (Qwen)
* **Perplexity Sonar Pro** (Perplexity)
### Standard (non-rate-limited) Models
Standard models do not count against your premium quota and are available in unlimited quantities on all paid plans. They include:
* GPT-5-mini, GPT-4o-mini, GPT-4.1 (OpenAI)
* Claude Haiku 4.5 (Anthropic)
* Gemini 2.5 Flash, Gemini Flash Lite (Google)
* DeepSeek Chat, DeepSeek Coder (DeepSeek)
* Grok-3, Grok Code Fast (xAI)
* Mistral Small, Magistral, Nemo (Mistral)
* Command R, Command R+, Command R7b (Cohere)
* Qwen Plus, Qwen Turbo, Qwen Flash, Qwen3 variants (Qwen)
* All Groq-hosted models
* All Kimi K2, Venice, TheSys, ZAI, Inception, ByteDance models
### Fallback Models
When your premium quota is exhausted on Pro or Pro 2x, ZeroTwo automatically routes messages to a fallback standard model: GPT-5-mini, GPT-4o-mini, Gemini Flash Lite, Mistral Small, or Grok 4 Fast.
## Not Sure Which Model to Use?
**Start here:** GPT-5 and Claude Sonnet 4.6 are excellent all-rounders for most tasks. GPT-5 excels at coding and structured outputs; Claude Sonnet 4.6 excels at writing, analysis, and nuanced tasks. Both have strong context windows (128k and 200k tokens respectively).
Claude Sonnet 4.6 and Claude Opus 4.6 are the strongest choices for nuanced long-form writing, document analysis, and tasks requiring careful reasoning about tone and context. Claude's 200k token context window makes it ideal for large documents. Gemini 2.5 Pro handles even longer documents (1M token context).
GPT-5 and DeepSeek Coder are strong for code generation, debugging, and refactoring. o3 and o4-mini are ideal for algorithmic problems requiring step-by-step reasoning. DeepSeek Coder is a standard model — excellent for developers on Free or Pro plans who want to preserve premium quota.
Perplexity Sonar Pro and Sonar are built for web-grounded answers with citations. Alternatively, enable Web Search with any tool-capable model (GPT-5, Claude Sonnet 4.6, Gemini 2.5 Pro) to ground responses in live data.
Groq-hosted models offer ultra-low latency. Gemini Flash Lite, GPT-4o-mini, and Mistral Small are fast standard models suitable for quick iteration, brainstorming, and high-volume tasks.
Qwen models (Qwen Max, Qwen3) have strong Chinese-language and multilingual capabilities. Mistral models perform well across European languages. Claude and GPT-5 handle a broad range of languages well.
Gemini 2.5 Pro with its 1M token context window is the best choice for extremely long documents, large codebases, or extended research sessions. Claude Sonnet 4.6 (200k tokens) is a strong alternative with excellent comprehension.
## Related Pages
* [Model Picker](/overview/model-picker) — how to select, search, and switch models in the chat interface
* [Plans and Pricing](/overview/plans-and-pricing) — which models count as premium and quota details
* [Plan Availability Matrix](/overview/plan-availability-matrix) — full feature-by-plan comparison
* [Answer Quality and Limitations](/core-chat/answer-quality-and-limitations) — choosing the right model for accuracy