Yakitori
Open-source AI. Uncensored by design.
Free tier · No credit card · Pro $18/mo
Capabilities
One studio for chat, image, and video.
Every modality, every model, every parameter — in one private workspace. No platform switching.
Frontier 70B models without filters
Dolphin, Hermes, Qwen — pick the right open-source model for the job. Stream replies, rewrite, regenerate, edit prompts mid-conversation.
- Llama-3 70B, Qwen2-72B class models
- Markdown + syntax-highlighted code
- Edit & regenerate any message
FLUX-quality images at any aspect
FLUX.1 Schnell for fast ideation, FLUX.1 Dev for finals. Up to 4 images per generation, every parameter tweakable.
- FLUX.1 Schnell · FLUX.1 Dev
- 1:1, 16:9, 9:16, 4:3, 3:4
- Seed, guidance, steps fully exposed
Open-source video that runs
Wan 2.1 T2V on RunPod Serverless. 2–8 seconds at 480p–720p. Queue jobs, cancel, download — never lose the prompt.
- Wan 2.1 T2V 14B
- Text-to-video, 16:9 / 9:16 / 1:1
- Async queue with progress
Privacy
Open AI doesn't have to mean open data.
The reason most people give up on local AI is the hardware. The reason most people don't trust hosted AI is the data policies. Yakitori removes both — we host the GPUs, you keep the data.
Stays in your browser
Conversations and generations persist in IndexedDB. They never leave the device unless you explicitly export.
No training on your data
Prompts are routed to GPUs for inference and discarded. We don't train on them, sell them, or share them.
No platform refusals
Open-source models. No mandatory safety layer. You decide what to ask and what to make.
Pseudonymous by default
Free tier needs no email. Pro uses minimal auth. We never see what you typed.
Models
Frontier open-source. Curated, not endless.
We don't throw 500 models at you. We pick the best uncensored open-source release in each modality and keep them current.
| Model | Provider | Modality | Notes | Plan |
|---|---|---|---|---|
| Dolphin 2.9 · Llama 3 8B | Cognitive Computations | Chat | 8K context · fast | Free |
| Hermes 3 · 70B | Nous Research | Chat | 32K context · reasoning | Pro |
| Dolphin 2.9.2 · Qwen2 72B | Cognitive Computations | Chat | 32K · code & multilingual | Pro |
| FLUX.1 Schnell | Black Forest Labs | Image | 4 steps · 1MP | Free |
| FLUX.1 Dev | Black Forest Labs | Image | 25 steps · high-fidelity | Pro |
| Wan 2.1 T2V · 14B | Alibaba | Video | 5s · up to 720p | Pro |
New models added monthly. Request the one you want in the Pro plan.
Pricing
Pick your plan. Cancel anytime.
Free is generous. Pro unlocks everything. Pro+ and Max scale with your usage. Billed in USDC on Base, no card needed.
Free
Base AI models
Explore Yakitori with the base model.
Start free- Dolphin Llama 3 8B (base)
- 10 text prompts per day
- 5 image prompts per day
- Private & uncensored
- Browser-local history
Pro
Full access
Your private AI studio — every model, every medium, no limits.
- All Pro models (Qwen3 14B Q4_K_M, AEON 27B)
- Unlimited text prompts
- 200 images per day (FLUX Schnell & Dev)
- Extended context windows up to 32K
- Priority compute · faster cold starts
- Encrypted chat backup & restore
- 100 credits / month for video & frontier models
Pro+
Scaled up
Everything in Pro, scaled for serious creators and users.
Go Pro+- Everything in Pro
- Higher image generation limits
- 7,500 credits / month for video, music, frontier models, and API
- 2-month credit banking — unused credits roll forward
- Higher rate limits across all modalities
Max
No ceilings
For the ultimate access for Yakitori power users.
Go Max- Everything in Pro+
- Highest image generation limits
- 22,500 credits / month
- 3-month credit banking
- Priority queue · first access to new models
USDC subscription on Base · No card needed · Cancel any time · Taxes calculated at checkout
FAQ
Answers, not asterisks.
The AI you wished
you could run locally.
Without the hardware. Without the filters.