Any model · one work layer
Use any model. Keep the same work.
NOME routes to the model you choose — frontier cloud or on-device — while your memory, context, approvals, artifacts, and receipts stay the same. The model can change; the work survives.
OpenAI
GPT · Codex
Anthropic
Claude
Google Gemini
Gemini · Vertex
DeepSeek
DeepSeek V4
xAI Grok
Grok
Meta Llama
Llama
Mistral
Mistral
MiniMax
MiniMax
NOME MLX local
On-device · offline
Bring your own provider subscription, use NOME’s hosted lane, or run on-device. Model availability depends on your connected providers.
How routing works
1 · Choose the model
Pick any provider per task — or let NOME pick. Switch models mid-thread without losing context.
2 · Choose the host & auth
Run on your own provider keys, NOME's hosted lane, or on-device MLX. Your data, your auth lane.
3 · Keep the work
Memory, context, approvals, artifacts, and receipts are the same no matter which model answered.