> ## 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. # Pricing ## Simple pricing for everyone Each paid plan includes a monthly allowance of free Auto requests. Models, images, video, and deep research share one credit pool. Top up anytime — purchased credits never expire. ## Plans Perfect for getting started. * 15 messages per day * 1 file upload per message * Limited model selection * Limited image generations, projects, tools, and memories * Serper web search For everyday AI use (and RP). Annual: \$11.99/mo. * 1,000 free Auto requests / month * 1,650 monthly credits * Use credits across all models and tools * Image generation and deep research * Memory unlocked * Unlimited projects * Multiple file uploads per task * Pets * Exa web search For professionals and creators. Annual: \$26.99/mo. * 2,500 free Auto requests / month * 3,500 monthly credits * Unlimited fast models in chat * Expanded image model selection * Expanded memory * Deep research, ZeroCode, Work agent, and video generation * Pets * Exa, Perplexity, and Google web search Double Pro usage. Annual: \$53.99/mo. * 5,000 free Auto requests / month * 7,000 monthly credits * Everything in Pro at higher limits Maximum power for power users. Annual: \$96/mo (\$1,152/year — 20% off). * 12,000 free Auto requests / month * 14,000 monthly credits * Unlimited fast models in chat * Highest individual credit allowance * Priority processing and faster responses * Maximum memory * Image, video, and deep research included * ZeroCode and Work agent * Early access to new features For roleplay, characters, and uncensored chat. Annual: \$7.99/mo. * 500 free Auto requests / month * 1,000 monthly credits * Unlimited character creation * Uncensored and roleplay model access * Character image generation * Pets * Access to regular chat models * Expanded memory Perfect for teams and organizations. Annual: \$35.99 / seat / month. * 4,000 monthly credits per seat (pooled) * 1,250 free Auto requests / seat / month (pooled) * Pooled team credits and shared workspaces * High-quality images at any scale * Maximum memory * Deep research, workflows, video, and shared projects * Simplified billing and user management * Early access to experimental features Team and Enterprise options are available for larger organizations — contact support at [reed@zerotwo.ai](mailto:reed@zerotwo.ai). ## How credits work Every model, image, video, and deep-research run draws from the same monthly credit pool. Heavier models and longer tasks use more credits; light chat uses very few. * **1 credit = \$0.01** — the single unit every model, tool, and generation is billed in * Paid plans include free **Auto** requests each month; after that, Auto falls back to credits * Run out? **Top up anytime** — purchased credits never expire ### Credit top-ups | Pack | Price | Credits | | ------- | ----- | ------- | | Starter | \$10 | 4,000 | | Builder | \$25 | 11,000 | | Power | \$50 | 24,000 | | Ultra | \$100 | 50,000 | Business workspaces can buy team top-up packs that apply to the pooled org wallet: | Pack | Price | Credits | | ------ | ----- | ------- | | Team | \$50 | 27,500 | | Growth | \$100 | 60,000 | | Scale | \$250 | 165,000 | ## Model and tool pricing Every rate on this page is in **credits** (1 credit = \$0.01). Token rates are per **1M tokens**, and each request is rounded up to the next whole credit. Rates track upstream provider pricing and change when providers change theirs. A typical chat turn is a few thousand tokens. At 200 credits / 1M input and 1,200 credits / 1M output, a 2,000-token prompt with a 500-token reply on **GPT 5.6 Luna** costs 1 credit. ### Chat and language models **Input** is what you send (prompt, files, tool results). **Output** is what the model writes back, including reasoning tokens. **Cached input** is the discounted rate for prompt-cache hits — long threads and reused system prompts hit the cache often, so real-world input cost usually lands below the full rate. A dash means the provider does not bill cached input separately. Models marked \* have long-context tiers — see [Long-context tiers](#long-context-tiers) below. | Model | Input | Cached input | Output | | ---------------- | ----- | ------------ | ------ | | GPT 5.6 Sol \* | 1,000 | 100 | 6,000 | | GPT 5.6 Terra \* | 500 | 50 | 3,000 | | GPT 5.6 Luna \* | 200 | 20 | 1,200 | | GPT 5.5 \* | 1,000 | 100 | 6,000 | | GPT 5.4 \* | 500 | 50 | 3,000 | | GPT 5.4 Mini | 150 | 15 | 900 | | GPT 5.4 Nano | 40 | 4 | 250 | | GPT 5.3 Chat | 500 | 50 | 3,000 | | GPT 5.3 Codex | 350 | 35 | 2,800 | | GPT 5.2 | 500 | 50 | 3,000 | | GPT 5.2 Codex | 350 | 35 | 2,800 | | GPT 5.1 | 500 | 50 | 3,000 | | GPT 5.1 Codex | 500 | 50 | 3,000 | | GPT 5 | 500 | 50 | 3,000 | | GPT 5 Mini | 50 | 5 | 400 | | GPT 5 Nano | 10 | 1 | 80 | | GPT 4.1 | 400 | 40 | 1,600 | | GPT 4.1 Mini | 80 | 8 | 320 | | GPT 4.1 Nano | 20 | 2 | 80 | | GPT 4o | 500 | 50 | 2,000 | | GPT 4o Mini | 30 | 3 | 120 | | o3 | 400 | 40 | 1,600 | | o4 Mini | 220 | 22 | 880 | | Model | Input | Cached input | Output | | ----------------- | ----- | ------------ | ------ | | Claude Opus 5 | 1,000 | 100 | 5,000 | | Claude Sonnet 5 | 400 | 40 | 2,000 | | Claude Opus 4.8 | 1,000 | 100 | 5,000 | | Claude Opus 4.7 | 1,000 | 100 | 5,000 | | Claude Opus 4.6 | 1,000 | 100 | 5,000 | | Claude Sonnet 4.6 | 600 | 60 | 3,000 | | Claude Sonnet 4.5 | 600 | 60 | 3,000 | | Claude Haiku 4.5 | 200 | 20 | 1,000 | Anthropic also bills **cache writes** at 1.25× the input rate when a prompt is first written to the cache. | Model | Input | Cached input | Output | | --------------------- | ----- | ------------ | ------ | | Gemini 3.1 Pro \* | 250 | 25 | 2,000 | | Gemini 3.5 Flash | 300 | 30 | 1,800 | | Gemini 3 Flash | 100 | 10 | 600 | | Gemini 3.1 Flash-Lite | 20 | 2 | 80 | | Gemini 2.5 Pro \* | 250 | 25 | 2,000 | | Gemini 2.5 Flash | 60 | 6 | 500 | | Gemini 2.5 Flash Lite | 20 | 2 | 80 | | Model | Input | Cached input | Output | | ------------------------ | ----- | ------------ | ------ | | Grok 4.5 | 400 | 100 | 1,200 | | Grok 4.3 | 250 | 40 | 500 | | Grok 4.2 \* | 250 | 40 | 500 | | Grok 4.2 R \* | 250 | 40 | 500 | | Grok 4 | 600 | 150 | 3,000 | | Grok 4 Fast \* | 40 | 10 | 100 | | Grok 4 Fast Reasoning \* | 40 | 10 | 100 | | Grok 4.1 Fast | 40 | 10 | 100 | | Grok 4.1 Fast R | 40 | 10 | 100 | | Grok Build 0.1 | 200 | 40 | 400 | | Grok Code Fast | 40 | 4 | 300 | | Model | Input | Cached input | Output | | -------------------- | ----- | ------------ | ------ | | Qwen 3.8 Max | 400 | 50 | 1,200 | | Qwen 3.7 Max | 500 | — | 1,500 | | Qwen 3.7 Plus | 80 | — | 320 | | Qwen 3.6 Plus | 100 | — | 600 | | Qwen 3.6 Flash | 38 | — | 226 | | Qwen 3.5 Plus | 80 | — | 480 | | Qwen3 Coder Plus \* | 200 | 40 | 1,000 | | Qwen3 Coder Flash \* | 60 | — | 300 | | Qwen3 Next 80B I | 30 | — | 184 | | Qwen3 Next 80B T | 30 | — | 240 | | Qwen Plus Character | 100 | — | 400 | | Qwen Flash Character | 10 | — | 80 | | Model | Input | Cached input | Output | | -------------- | ----- | ------------ | ------ | | Kimi K3 | 600 | 60 | 3,000 | | Kimi K2.7 Code | 190 | 38 | 800 | | Kimi K2.6 | 190 | 32 | 800 | | Kimi K2.5 | 120 | 20 | 600 | | Model | Input | Cached input | Output | | ----------------- | ----- | ------------ | ------ | | DeepSeek V4 Pro | 87 | 0.7 | 174 | | DeepSeek V4 Flash | 28 | 0.6 | 56 | | Model | Input | Cached input | Output | | --------------- | ----- | ------------ | ------ | | GLM 5.2 | 280 | 52 | 880 | | GLM 5.1 | 350 | 66 | 1,100 | | GLM 5 | 200 | 22 | 640 | | GLM 4.7 | 110 | 22 | 530 | | GLM 4.7 Flash | 14 | — | 80 | | GLM 4.7 Flash H | 28 | — | 160 | | GLM 4.6 | 120 | 10 | 440 | | Model | Input | Cached input | Output | | ------------- | ----- | ------------ | ------ | | MiniMax M3 \* | 60 | 12 | 240 | | MiniMax M2.7 | 60 | 12 | 240 | | MiniMax M2.5 | 60 | 6 | 240 | | Model | Input | Cached input | Output | | ------------------- | ----- | ------------ | ------ | | Command A+ | 500 | — | 2,000 | | Command A | 500 | — | 2,000 | | Command A Reasoning | 500 | — | 2,000 | | Command R7B | 7.5 | — | 30 | | Model | Input | Cached input | Output | | --------- | ----- | ------------ | ------ | | Sonar Pro | 600 | — | 3,000 | | Sonar | 200 | — | 200 | | Model | Input | Cached input | Output | | -------------------------- | ----- | ------------ | ------ | | Venice 1.2 RP | 100 | — | 400 | | Venice 1.2 | 40 | — | 180 | | Qwen 3.6 Plus (uncensored) | 125 | 12.5 | 750 | | Qwen3 235B T | 90 | — | 700 | | Gemma 4 (uncensored) | 32 | — | 100 | | Gemma 4 31B | 24 | 18 | 72 | | Mistral 3.1 24B | 6 | — | 22 | | Model | Provider | Input | Cached input | Output | | ---------------------------- | ----------- | ----- | ------------ | ------ | | Mistral Small | Mistral | 30 | 3 | 120 | | C1/Sonnet 4 | Thesys | 600 | 60 | 3,000 | | C1/GPT-5 | Thesys | 250 | 25 | 2,100 | | Muse Spark 1.2 | Meta | 250 | 30 | 850 | | Muse Spark 1.2 (Contributor) | Meta | 20 | 0.4 | 40 | | Muse Spark 1.1 | Meta | 250 | 30 | 850 | | Mercury 2 | Inception | 50 | 5 | 150 | | Gemma 4 31B | Gemma | 24 | 18 | 72 | | MiMo V2.5 Pro | MiMo | 87 | 0.7 | 174 | | MiMo V2.5 | MiMo | 28 | 0.6 | 56 | | GPT-OSS 120B | Fireworks | 30 | 3 | 120 | | GPT-OSS 20B | Fireworks | 14 | 7 | 60 | | Inkling | Together AI | 200 | 34 | 810 | #### Long-context tiers Some providers charge a higher rate once a single request's input passes a threshold. The higher rate applies to the **whole request**, not just the tokens above the line. | Model | Threshold | Input above | Cached input above | Output above | | ---------------------------- | ----------------- | ----------- | ------------------ | ------------ | | GPT 5.6 Sol | 272K input tokens | 2,000 | 200 | 9,000 | | GPT 5.6 Terra | 272K | 1,000 | 100 | 4,500 | | GPT 5.6 Luna | 272K | 400 | 40 | 1,800 | | GPT 5.5 | 272K | 2,000 | 200 | 9,000 | | GPT 5.4 | 272K | 1,000 | 100 | 4,500 | | Gemini 3.1 Pro | 200K | 500 | 50 | 3,000 | | Gemini 2.5 Pro | 200K | 500 | 50 | 3,000 | | Grok 4.2 / 4.2 R | 200K | 400 | 100 | 1,200 | | Grok 4 Fast / Fast Reasoning | 128K | 80 | — | 200 | | MiniMax M3 | 512K | 120 | 24 | 480 | Qwen's coder models use four brackets instead of two: | Model | ≤32K | ≤128K | ≤256K | ≤1M | | ----------------- | ----------- | ----------- | ----------- | -------------- | | Qwen3 Coder Plus | 200 / 1,000 | 360 / 1,800 | 600 / 3,000 | 1,200 / 12,000 | | Qwen3 Coder Flash | 60 / 300 | 100 / 500 | 160 / 800 | 320 / 1,920 | Values are **input / output** credits per 1M tokens. ### Auto **Auto** picks a fast, capable model for you. Auto requests are **free** up to your plan's monthly allowance; after that they bill at the target model's normal rate — **GPT 5.4 Nano** in chat, **MiniMax M3** in Work, Cowork, ZeroCode, and deep research. | Plan | Free Auto requests / month | | ---------- | -------------------------------------------- | | Free | Counted against the daily free-message limit | | Roleplay | 500 | | Plus | 1,000 | | Pro | 2,500 | | Pro 2× | 5,000 | | Plus Ultra | 12,000 | | Business | 1,250 per seat (pooled) | ### Image models Credits per image at the default quality and a 1024² baseline. Higher quality tiers and larger resolutions cost more; requesting several images multiplies the charge. | Model | Provider | Modes | Credits / image | | ------------------------------------ | ----------------- | -------------- | --------------- | | Nano Banana Pro | Google | Generate, edit | 30 | | Nano Banana 2 | Google | Generate, edit | 16 | | Qwen Image 2 Pro | Qwen | Generate, edit | 15 | | Ideogram 2.0 | Ideogram | Generate | 12 | | Flux Pro | Black Forest Labs | Generate | 11 | | GPT Image 2 | OpenAI | Edit | 10.6 | | Imagen 4 | Google | Generate | 8 | | Qwen Image 3 | Qwen | Generate, edit | 8 | | Nano Banana (Gemini 2.5 Flash Image) | Google | Generate, edit | 7.8 | | Qwen Image 2 | Qwen | Generate, edit | 7 | | Seedream v5 Lite | ByteDance | Generate, edit | 7 | | GPT Image 1.5 | OpenAI | Generate, edit | 6.8 | | GPT Image 1 | OpenAI | Generate, edit | 6.8 | | Flux Pro 2 | Black Forest Labs | Generate, edit | 6 | | Qwen Edit | Qwen | Edit | 6 | | Imagen 4 Fast | Google | Generate | 4 | | Grok Imagine | xAI | Generate, edit | 4 | | Grok Imagine Pro | xAI | Generate, edit | 4 | | Qwen | Qwen | Generate | 4 | | GPT Image 1 Mini | OpenAI | Generate, edit | 2.2 | | Z-Image Turbo | Tongyi | Generate | 1 | **GPT Image 1 Mini** is the default image model and the only one available on Free. Plus adds **Imagen 4 Fast**, **Seedream v5 Lite**, and **Z-Image Turbo** for generation. Pro and above unlock the full list. See [Image generation](/image-generation). ### Video models Video is billed on the **generation provider's reported cost** for the finished clip. When the provider does not return a cost, it falls back to **10 credits per second** of output. | Model | Provider | Duration | Notes | | ------------------------ | --------- | -------- | ---------------------------------------------- | | Seedance 1.5 Pro | ByteDance | 4–12s | Text-to-video and image-to-video, 720p default | | Kling VIDEO 3.0 Standard | Kling AI | 3–15s | Text-to-video and image-to-video | | Grok Imagine Video | xAI | 1–15s | Text-to-video and image-to-video | | Wan2.6 Flash | Alibaba | 2–15s | Image-to-video only | Longer clips and higher resolutions cost proportionally more. Video generation is not available on Free. ### Audio models | Type | Model | Rate | | ------------------ | --------------------------- | ------------------------------- | | Speech and podcast | ElevenLabs v3 | 36 credits per 1,000 characters | | Sound effects | ElevenLabs Text-to-Sound v2 | 1.6 credits / second | | Music | Gemini Lyria | 0.3 credits / second | Music clips run 5–60 seconds (30s default), so a default track is about 9 credits. Sound effects run 0.5–30 seconds. Saved voice management — creating, listing, and deleting voice profiles — is free. ### Tool calls Tools that hit a paid third-party API are charged individually and appear as their own line in **Credit Activity**. Each priced call is rounded up to the next whole credit. | Tool | Provider | Credits | | ----------------------- | --------------------------------- | ------------------------------- | | Web search | Exa (default) | Exa's reported cost per search | | Web search | Serper | 1 per search | | Web search — Google | SerpApi | 5 per search | | Web search | Perplexity Sonar Pro | Billed at Sonar Pro token rates | | Image search | SerpApi | 5 per search | | X search | xAI Grok | xAI's reported cost per request | | Web page fetch | Firecrawl | 2 per page | | Image generation / edit | See [Image models](#image-models) | Per image | | Video generation | See [Video models](#video-models) | Per clip | | Audio generation | See [Audio models](#audio-models) | Per clip | Search engine availability depends on your plan: Free uses Serper, Plus adds Exa, and Pro and above add Perplexity and Google. Everything else — file reads and writes, code execution, bash, memory, planning, connectors, and MCP tool calls — adds **no tool cost**. You pay only for the model tokens those calls consume, and they are still logged at 0 credits so you can see every step. ### Bring your own subscription Models you connect on the desktop app through your own provider subscription or API key — ChatGPT, Claude, Grok, Qwen, MiniMax, GitHub Copilot, OpenRouter, Kimi Code — run against **that account**, not your ZeroTwo credits. They are billed at 0 credits here. See [Models](/models#bring-your-own-subscription-desktop). ## Related * [Models](/models) * [Quickstart](/quickstart) * [Open ZeroTwo](https://zerotwo.ai)