# PromptKit
> Expert-crafted system prompts for 200+ AI models. Copy, paste, and get optimized outputs for ChatGPT, Claude, Gemini, Midjourney, and more. System prompts are derived from official documentation, whitepapers, and model cards.

## Text & Chat Models
- [GPT-5.5](https://promptkit.qzz.io/model/sp-gpt55) — OpenAI's newest frontier model. Outcome-first prompting, reasoning effort control. | [GPT-5](https://promptkit.qzz.io/model/sp-gpt5) | [GPT-4o](https://promptkit.qzz.io/model/sp-gpt-4o) | [o3](https://promptkit.qzz.io/model/sp-o3) | [o4-mini](https://promptkit.qzz.io/model/sp-o4-mini)
- [Claude Opus 4](https://promptkit.qzz.io/model/sp-claude-opus-4) — Anthropic's most capable model. XML system messages, thinking mode. | [Claude Sonnet 4](https://promptkit.qzz.io/model/sp-claude-sonnet-4) | [Claude Haiku 4](https://promptkit.qzz.io/model/sp-claude-haiku-4) | [Claude Fable 5](https://promptkit.qzz.io/model/sp-claude-fable-5)
- [Gemini 3.5](https://promptkit.qzz.io/model/sp-gemini-35) — Google's latest. System instructions, Search grounding, thinking config. | [Gemini 2.5 Pro](https://promptkit.qzz.io/model/sp-gemini-25-pro) | [Gemma 4](https://promptkit.qzz.io/model/sp-gemma-4)
- [DeepSeek R1](https://promptkit.qzz.io/model/sp-deepseek-r1) — Reasoning mode, temperature ranges, tool use. | [DeepSeek V4](https://promptkit.qzz.io/model/sp-deepseek-v4)
- [Qwen3.7](https://promptkit.qzz.io/model/sp-qwen37) — Chat template format, tool descriptions. | [QwQ](https://promptkit.qzz.io/model/sp-qwq) | [Qwen Max](https://promptkit.qzz.io/model/sp-qwen-max)
- [Llama 4](https://promptkit.qzz.io/model/sp-llama-4) — System prompt structure, tool use format. | [Llama 3](https://promptkit.qzz.io/model/sp-llama-3)
- [Mistral Large](https://promptkit.qzz.io/model/sp-mistral-large) — System message guidelines, JSON mode. | [Mistral Small](https://promptkit.qzz.io/model/sp-mistral-small)
- [GLM-5](https://promptkit.qzz.io/model/sp-glm5) — Chinese-optimized prompt formats. | [GLM-4](https://promptkit.qzz.io/model/sp-glm-4)
- [Doubao](https://promptkit.qzz.io/model/sp-doubao) — ByteDance's Doubao series. | [Grok 3](https://promptkit.qzz.io/model/sp-grok-3) — xAI's reasoning model.
- [Kimi K2](https://promptkit.qzz.io/model/sp-kimi-k2) — Moonshot AI reasoning, K2.5 thinking mode. | [MiniMax M3](https://promptkit.qzz.io/model/sp-minimax-m3)
- [Command A](https://promptkit.qzz.io/model/sp-command-a) — Cohere's Command A7, Command R+. | [Yi-Lightning](https://promptkit.qzz.io/model/sp-yi-lightning) — 01.AI's fast model.
- [Mercury Coder](https://promptkit.qzz.io/model/sp-mercury-coder) — Inception Labs diffusion LLM.

## Image Generation
- [Midjourney V7](https://promptkit.qzz.io/model/sp-midjourney-v7) — Parameters (--v, --s, --p, --ar), raw mode, personalization
- [Flux 2](https://promptkit.qzz.io/model/sp-flux-2) — Black Forest Labs. Guidance, aspect ratio, LoRA, Kontext editing. | [Flux Pro](https://promptkit.qzz.io/model/sp-flux-pro) | [Flux Dev](https://promptkit.qzz.io/model/sp-flux-dev)
- [SDXL](https://promptkit.qzz.io/model/sp-sdxl) — Stability AI. Community quality tags, negative prompts, CFG scale. | [SD 3.5](https://promptkit.qzz.io/model/sp-sd-35)
- [GPT Image 2](https://promptkit.qzz.io/model/sp-gpt-image-2) — OpenAI. Multi-turn editing, style presets. | [DALL-E 3](https://promptkit.qzz.io/model/sp-dall-e-3)
- [Ideogram V4](https://promptkit.qzz.io/model/sp-ideogram-v4) — Text rendering, magic prompt. | [Recraft V3](https://promptkit.qzz.io/model/sp-recraft-v3) — Style control, vector output.
- [Adobe Firefly 3](https://promptkit.qzz.io/model/sp-adobe-firefly-3) — Commercial safety. | [Imagen 3](https://promptkit.qzz.io/model/sp-imagen-3) — Google's image model.

## Video Generation
- [Sora 2](https://promptkit.qzz.io/model/sp-sora-2) — OpenAI. Text generation, camera motion, video-to-video
- [Kling 3](https://promptkit.qzz.io/model/sp-kling-3) — Kuaishou. Text-to-video, image-to-video, motion dynamics
- [Runway Gen 4](https://promptkit.qzz.io/model/sp-runway-gen-4) — Ultra-realistic video generation
- [Pika 3.0](https://promptkit.qzz.io/model/sp-pika-3) — Scene ingredients, Pikaffects
- [Veo 3.1](https://promptkit.qzz.io/model/sp-veo-31) — Google DeepMind. Audio-capable video generation

## Vision & Multimodal
- [GPT-5 Vision](https://promptkit.qzz.io/model/sp-gpt5-vision) — Image analysis, structured outputs from visuals
- [Claude Vision](https://promptkit.qzz.io/model/sp-claude-vision) — Image understanding, PDF analysis
- [Gemini Vision](https://promptkit.qzz.io/model/sp-gemini-vision) — Video analysis, image understanding

## 3D Generation
- [Tripo 3D](https://promptkit.qzz.io/model/sp-tripo-3d) — Text-to-3D, image-to-3D
- [Meshy 3D](https://promptkit.qzz.io/model/sp-meshy-3d) — Text-to-3D, PBR textures

## Open-Weight Models
- [Flux Dev](https://promptkit.qzz.io/model/sp-flux-dev) — Black Forest Labs open model
- [Llama 4](https://promptkit.qzz.io/model/sp-llama-4) — Meta's open model
- [Gemma 4](https://promptkit.qzz.io/model/sp-gemma-4) — Google's open model
- [DeepSeek R1](https://promptkit.qzz.io/model/sp-deepseek-r1) — Open reasoning model

## API Reference
- Site: https://promptkit.qzz.io
- Sitemap: https://promptkit.qzz.io/sitemap.xml
- 401 model pages with unique system prompts
- Each page contains: system prompt, pro tips, source references, arena rankings