DarkGPT
How to choose a model
- For quick, simple questions: the small models (Mistral 3B, Mistral 14B, Qwen 3 235B) cost very little.
- For code: Kimi K2.7 Code, Devstral 2, Codestral, or Claude Sonnet 4.6 at the high end. See AI for coding.
- For writing: Claude Sonnet 4.6, MiniMax M3, ChatGPT 4.1.
- For long documents: Gemini 2.5 Pro, Llama 4 Scout, MiniMax 2.7.
- For images and screenshots: the models with vision (Gemini, GLM 4.5V, Llama 4 Maverick).
- For hard problems: Claude Opus, DeepSeek v4 Pro, with Thinking mode.
You can switch models at any time, even in the middle of a conversation.
Unfiltered mode, model by model
- Grok 4.2Dark humor, blunt feedback, firm opinions.
- DeepSeek v4 ProHistory, debates and touchy topics, no hedging.
- Kimi K2.5Dark fiction and long scenes, no softening.
- Qwen 3 235B (free)Frank everyday questions, for 15 gems.
Which plan unlocks what?
| Plan | Models | Price |
|---|---|---|
| Free | 5 models: Mistral 3B, Qwen 3 235B, Mistral Small / 12B, Mistral 14B, DeepSeek v4 Flash | €0 |
| Basic | 35 models: everything except Claude and ChatGPT 4.1, with OSINT, images, retouching and DocGPT | €21.99/month |
| Gold | All 42: Claude Sonnet, Claude Opus and ChatGPT 4.1 included, with DreamGPT Video | €51.99/month |
| Platinum and Diamond | All 42, with more gems per day and Thinking mode | from €89.99/month |
Unfiltered mode: which models answer straight
Most of the 42 models carry the Unfiltered mode tag in the catalog: their filters are kept to a minimum, and they answer legitimate adult requests without moralizing.
That covers the Mistral, Qwen, DeepSeek, Gemini, Grok, GLM, Kimi and Llama families. Claude and ChatGPT stay cautious: keep them for writing quality or code, and switch to an Unfiltered mode model when they dodge, in the same conversation.
For concrete examples, see our 40 uncensored prompts. Adults 18+ only; illegal use is banned.
The full catalog
Gem price per message, Thinking mode extra where available, and minimum plan.
Fast and budget-friendly 12
| Model | Gems | Plan |
|---|---|---|
M Mistral 3BMistral · Ultra-light: near-instant answers for simple questions. | 5 | Free |
A Qwen 3 235BAlibaba · Alibaba’s open-source giant, surprisingly strong for its price. | 15+20 think | Free |
M Mistral Small / 12BMistral · The sweet spot between speed and quality, made in France. | 20 | Free |
M Mistral 14BMistral · Versatile and fast, perfect for everyday tasks and translation. | 30 | Free |
D DeepSeek v4 FlashDeepSeek · All of DeepSeek’s logic in express mode, very good at code. | 50+20 think | Free |
Z GLM 4.5 AirZ.ai · The lightweight GLM: fast and reliable. | 70 | Basic+ |
G Gemini 2.5 Flash VisionGoogle · Fast and multimodal, ideal for anything that needs to move quickly. | 80+30 think | Basic+ |
x Grok 3 MinixAI · Small but sharp: reasons fast on short questions. | 80 | Basic+ |
O ChatGPT 3.5 TurboOpenAI · The model that started it all: simple, fast and effective. | 90 | Basic+ |
G Gemini 3.1 Flash Lite VisionGoogle · Next-gen Gemini in a lightweight version. | 90+20 think | Basic+ |
x Grok 4.1 FastxAI · Grok in fast mode: direct, no detours. | 100 | Basic+ |
G Gemini 3.5 Flash Lite VisionGoogle · The latest Gemini generation, light and responsive. | 110+40 think | Basic+ |
Coding specialists 5
| Model | Gems | Plan |
|---|---|---|
M CodestralMistral · Mistral’s code specialist: completion, refactoring, scripts. | 75 | Basic+ |
M Devstral Small 2Mistral · Built for development: reads, understands and edits your code. | 80 | Basic+ |
M Kimi K2Moonshot · Excellent at code and at long, structured tasks. | 90 | Basic+ |
M Devstral 2Mistral · Mistral’s developer model, built for big projects. | 95 | Basic+ |
M Kimi K2.7 CodeMoonshot · Code specialist: big projects, debugging and refactoring. | 160 | Basic+ |
All-rounders and long context 11
| Model | Gems | Plan |
|---|---|---|
A Qwen 3 32BAlibaba · Compact and efficient, good at code and multilingual work. | 60+20 think | Basic+ |
M MiniMax 2.7MiniMax · Long context and fluid answers for big texts. | 70 | Basic+ |
M Llama 3.1 70BMeta · Meta’s open-source classic, robust and versatile. | 85 | Basic+ |
M Mistral MediumMistral · More depth than the small Mistrals, still fast. | 85 | Basic+ |
Z GLM 4.5V VisionZ.ai · GLM with vision: reads your images, screenshots and diagrams. | 90 | Basic+ |
M Llama 4 ScoutMeta · A giant context window for very long documents. | 90 | Basic+ |
M Kimi K2.5Moonshot · More accurate and more creative than K2. | 100 | Basic+ |
M Llama 4 Maverick VisionMeta · The most powerful Llama 4, multimodal and versatile. | 110 | Basic+ |
M MiniMax M3MiniMax · Next-gen MiniMax, very fluent at writing. | 120 | Basic+ |
G Gemini 2.5 VisionGoogle · Fully multimodal: text, images and documents in one conversation. | 120+60 think | Basic+ |
O ChatGPT 4.1OpenAI · OpenAI’s GPT, precise and versatile on every subject. | 220 | Gold+ |
Reasoning and flagship 9
| Model | Gems | Plan |
|---|---|---|
Z GLM 4.7Z.ai · The latest GLM, very strong at code and logic. | 110+35 think | Basic+ |
x Grok 4.2 Non ReasoningxAI · The latest Grok: blunt, quick and full of personality. | 120+50 think | Basic+ |
D DeepSeek v4 ProDeepSeek · Advanced reasoning: math, logic, reverse engineering and complex code. | 150+40 think | Basic+ |
G Gemini 2.5 Pro VisionGoogle · The most powerful Gemini, unbeatable on long documents. | 180+60 think | Basic+ |
A Claude Sonnet 4.5Anthropic · Ultra-reliable for long projects, code and writing. | 500+180 think | Gold+ |
A Claude Sonnet 4.6Anthropic · The flagship: unbeatable at code and writing. | 550+180 think | Gold+ |
A Claude Opus 4.5Anthropic · Deep reasoning for the hardest problems. | 650+220 think | Gold+ |
A Claude Opus 4.6Anthropic · The new Opus generation, faster and more precise. | 700+230 think | Gold+ |
A Claude Opus 4.7Anthropic · The most powerful model in the whole lineup. | 800+250 think | Gold+ |
DarkGPT tools 5
| Model | Gems | Plan |
|---|---|---|
D DarkGPT OSINTDarkGPT · Public-source research: usernames, domains, leaks, metadata. | 50 | Basic+ |
D DreamGPT ImageDarkGPT · Generates images from a simple description. | 100 | Basic+ |
D DreamGPT RetouchDarkGPT · Edits your images (background, objects, style) in one sentence. | 150 | Basic+ |
D DocGPTDarkGPT · Edits, summarizes and rewrites your documents directly. | 250 | Basic+ |
D DreamGPT VideoDarkGPT · Generates videos from text or an image. | 400 | Gold+ |
Frequently asked questions
How many models does DarkGPT offer?
42 models and tools from 11 providers, 5 of them free.
Which DarkGPT models are uncensored?
Most of them: the Mistral, Qwen, DeepSeek, Gemini, Grok, GLM, Kimi and Llama families run in Unfiltered mode. Claude and ChatGPT stay more cautious.
Which models are free?
Mistral 3B, Mistral Small / 12B, Mistral 14B, Qwen 3 235B and DeepSeek v4 Flash.
What is the most powerful model?
Claude Opus 4.7, the most powerful in the lineup (800 gems per message, from Gold).
Can I switch models during a conversation?
Yes, at any time, in one click.
