Full control over the model behind your chatbot.Powered by your provider.
Connect your own model from OpenAI, Anthropic, Google, DeepSeek, Qwen, Mistral, Moonshot, OpenRouter, Groq, or Requesty. Your provider, your model, your invoice — Gydr handles the harness.
How BYOM helps
Run on your model, keep the harness
All of Gydr's guardrails, retrieval, and skills — driven by the model you already pay for.
Connect your own provider
Ten providers supported: OpenAI, Anthropic, Google Gemini, DeepSeek, Qwen, Mistral, Moonshot, OpenRouter, Groq, and Requesty. One neutral interface.
Pay your provider directly
Chat credits are zero on Gydr while BYOM is active. You're billed by your provider. Retrieval, orchestration, embedding still consume Gydr credits.
Live import validation
Every imported model is validated against your key with a function-calling probe. Models that don't support tool calling are hard-blocked at import.
Per-widget failure behavior
Choose what happens if your provider has an outage: a graceful fallback message, or fall back to a Gydr model for that turn.
Standard Gydr models vs Bring Your Own Model
| Standard Gydr Models | Bring Your Own Model | |
|---|---|---|
| Inference billing | Per-turn Gydr credits | Your provider's billing |
| Available models | Curated Gydr set (Nova family + OSS) | Anything your provider supports |
| Provider choice | Bedrock-managed | OpenAI / Anthropic / Gemini / DeepSeek / Qwen / Mistral / Moonshot / OpenRouter / Groq / Requesty |
| Tool calling | Required | Required (validated on import) |
| Auxiliary models (planner, judge) | Always Bedrock | Always Bedrock |
| Failure handling | N/A | Graceful message or fall back to Gydr |
Supported providers
Proprietary models
OpenAI
Industry-leading models with reliable tool calling.
Anthropic
Claude family. Strong reasoning, transparent refusals.
Google Gemini
Long-context flagship models.
DeepSeek
Cost-efficient frontier models; strong at coding and math.
Qwen (Alibaba DashScope)
Alibaba's flagship; multilingual.
Mistral
Open-weight European provider; strong tool calling.
Moonshot (Kimi)
Ultra-long-context, agent-focused.
Aggregators
OpenRouter
Aggregator routing to most major models behind one key.
Groq
Hardware-accelerated inference for popular open models.
Requesty
Unified gateway with routing, caching, and one bill across 300+ models.
Three steps
Add a connection
In Console → Settings → Import Models, paste your provider API key. We encrypt it per-account and never show it again.
Import a model
Type the provider's model id. We run a live validation probe — only models with working tool calling are accepted.
Pick it on a widget
On any widget, choose your imported model under LLM Model and pick a failure behavior. Done.