Models
AI models for enterprise agents
Every model your agents call goes through the Model Broker. Pool providers for failover, decide who may call each model, and cap what it costs.
Models
AI models for enterprise agents
Any model your providers serve
Connect it, pool it for failover, and decide who may call it.
Claude FableModelsFable 5.1 from Anthropic and Amazon Bedrock in one failover pool, open to one group and capped by quota.Claude HaikuModelsHaiku 4.5 from Anthropic and Amazon Bedrock in one failover pool, as the default model for high-volume agents and contractors.Claude OpusModelsOpus 5.5 from Anthropic and AWS Bedrock in one failover pool, with rules on who calls it and quotas on what it costs.Claude SonnetModelsSonnet 5.5 from Anthropic and Amazon Bedrock in one failover pool, with agents pinned to it and quotas on spend.
DeepSeekModelsDeepSeek V4 Pro from Together AI and DeepInfra in one failover pool, with regulated work kept off DeepSeek's own API.
GeminiModelsGemini on Vertex AI behind the Model Broker: Flash with a fallback model, Pro kept to one group, spend capped by quota.
GemmaModelsGemma 4 31B from DeepInfra and NVIDIA NIM in one failover pool, with contractors kept on it and a cap on daily calls.
GPTModelsGPT-6 from OpenAI and Azure OpenAI in one failover pool, with rules on which deployment handles which work.gpt-ossModelsgpt-oss-120b from Groq, Cerebras and Together AI in one failover pool, with unattended jobs kept on open weights.
GrokModelsGrok 4.7 from xAI with OpenRouter as fallback in one pool, kept to the research team.
KimiModelsKimi K3 from Moonshot AI, Together AI and Fireworks AI in one failover pool, kept to the people who need it.
LlamaModelsLlama 3.3 70B from Groq, Together AI and SambaNova in one failover pool, limited to the teams that need it.
MistralModelsMistral Large 3 from Mistral and NVIDIA NIM in one failover pool, with EU work kept on EU connections.
QwenModelsQwen 3.8 from Together AI and DeepInfra in one failover pool, with an agent pinned to it and a cap on its spend.How it works
Any model, behind a pool
Connect a model from each provider, pool the connections, and agents send the pool's name. The list below is a convenience: any model id your provider accepts works.
01
Connect the model
One connection per provider and model. The provider key stays in Dome.
$ dome models add opus-55-anthropic \--provider anthropic --model claude-opus-5-5 \--api-key "$ANTHROPIC_API_KEY" --gateway prod-gateway02
Pool it for failover
Members are tried in priority order. Name the pool for the model.
$ dome models pool create claude-opus-5-5 --failover-max all --gateway prod-gateway$ dome models pool member add claude-opus-5-5 opus-55-anthropic --priority 0$ dome models pool member add claude-opus-5-5 opus-55-bedrock --priority 103
Send the pool's name
Agents call the Gateway with their own Dome key.
from openai import OpenAIclient = OpenAI(base_url=f"{GATEWAY_URL}/v1", api_key=DOME_AGENT_KEY)client.chat.completions.create(model="claude-opus-5-5", messages=messages)
Commands tested against a Dome workspace on September 30, 2026.
FAQ
Common questions
Can agents use a model that isn't listed?
Yes. Any model id your provider accepts works. The list is what the picker suggests.
How does failover work?
Pool members are tried in priority order. When one fails, the next attempt goes to the next member.
Can different agents get different models?
Yes. Rules decide which agents, and which people they act for, may call each pool or connection.
Next steps
Talk with our FDE team
Our forward deployed engineers work with your platform team to get your agents into production and under control: the first one governed on your own systems, and a pattern your teams can repeat for every agent after it.
No card required to start. Register your first agent in minutes.
LLM router guide
What an LLM router does, and what to ask of one before your agents depend on it.
Read the guide