The AI models that power your apps.
Pick the model that fits your need (from lightweight and instant to large and thorough) and give each application its own engine.
A model for every task.
- Complete control
Each application has its own model. Change it at any time without affecting anything else - no rebuild, no downtime.
- Private by design
All models run on dedicated infrastructure. Your prompts, instructions, and outputs stay in your workspace, never used to train anything.
- No API key needed
You pick the model; the platform handles the infrastructure, routing, and billing. Nothing to provision, nothing to manage.
Available models.
These open-source models are ready to use. Each one is pre-loaded, tested, and available from your app settings - just select and go. Every step of a workflow runs on the model of the app it calls, so a chained flow keeps the same engines you picked.
Qwen3
Alibaba
General-purpose model from the Qwen 3 family (Alibaba), balancing reasoning and speed.
Qwen3 Embedding
Alibaba
Embedding model from the Qwen 3 family (Alibaba), built for semantic search and vectorization.
Qwen3 VL
Alibaba
Vision-language model from the Qwen 3 family (Alibaba), built for image and document analysis.
Qwen3.5
Alibaba
Latest-generation general-purpose model from the Qwen family (Alibaba), powerful and versatile.
Code Gemma
Gemma variant (Google) specialized in code generation and completion.
Embedding Gemma
Compact Gemma-family embedding model (Google) for semantic search.
Gemma 3
Open general-purpose model from the Gemma 3 family (Google), multimodal and efficient.
Gemma 4
Open general-purpose model from the Gemma 4 family (Google), multimodal and available from compact to most powerful.
Shield Gemma
Gemma model (Google) dedicated to content moderation and safety.
Translate Gemma
Gemma-family model (Google) specialized in multilingual translation.
Codestral
Mistral AI
Mistral AI model specialized in code generation across many languages.
Devstral Small 2
Mistral AI
Devstral-family model (Mistral AI) specialized in software development and code generation.
Mathstral
Mistral AI
Mistral AI model specialized in mathematical and scientific reasoning.
Ministral 3
Mistral AI
Compact Mistral AI model designed for on-device efficiency and low latency.
Mistral
Mistral AI
General-purpose Mistral AI model known for its performance/efficiency balance.
Mistral Small 3.2
Mistral AI
Compact, fast Mistral AI model with a good quality/cost trade-off.
Granite 3.1 Dense
IBM
Dense model from the Granite 3.1 family (IBM), balanced for general tasks.
Granite 3.1 MoE
IBM
Mixture-of-experts (MoE) Granite 3.1 model (IBM), efficient at scale.
Granite 3.2 Vision
IBM
Compact model from the Granite 3.2 family (IBM) specialized in image and document analysis.
Granite 3.3
IBM
General-purpose model from the Granite 3 family (IBM), built for professional use.
Granite 4
IBM
Latest-generation general-purpose model from the Granite family (IBM), enterprise-oriented.
Granite 4.1
IBM
General-purpose model from the Granite 4.1 family (IBM), balanced for everyday text tasks.
Granite Code
IBM
Granite-family model (IBM) specialized in code generation.
Granite Guardian 4.1
IBM
Granite 4.1-family model (IBM) dedicated to content moderation and safety.
Phi 3.5
Microsoft
Capable small model from the Phi family (Microsoft), efficient for its size.
Phi 4
Microsoft
Compact model from the Phi 4 family (Microsoft), strong at reasoning for its size.
Phi 4 mini
Microsoft
Very small Phi 4 model (Microsoft), designed for speed and on-device use.
Phi 4 mini reasoning
Microsoft
Mini Phi 4 (Microsoft) specialized in low-cost reasoning.
Phi 4 reasoning
Microsoft
Phi 4 variant (Microsoft) optimized for step-by-step reasoning.
DeepSeek Coder V2
DeepSeek
DeepSeek model specialized in code generation and completion.
DeepSeek OCR
DeepSeek
Compact DeepSeek-family model (DeepSeek) specialized in document reading and text recognition.
DeepSeek R1
DeepSeek
DeepSeek reasoning model built for complex, step-by-step problem solving.
DeepSeek V2
DeepSeek
General-purpose mixture-of-experts model from DeepSeek, powerful and cost-efficient.
MiniCPM-V 4.5
OpenBMB
Compact multimodal model from the MiniCPM-V family (OpenBMB), efficient on both text and images.
MiniCPM-V 4.6
OpenBMB
Very compact multimodal model from the MiniCPM-V family (OpenBMB), built for lightweight processing.
LFM2.5
Liquid AI
General-purpose model from the LFM family (Liquid AI), compact and fast for text tasks.
Ready to build with the model of your choice?
Start free. Pick any model, write your instructions, and your app is ready in minutes.