Built-in providers
Every workspace includes Neutrome-managed model routes. You can use these model IDs immediately in an executor. You do not create a provider, configure an endpoint, or supply an upstream API key for them.
Use default/* for the normal managed route. Use fast/* only when the catalog offers the model and lower latency is more important for the request.
Prices below are mocked examples per 1M tokens. They are not billing rates and must not be used for cost estimation or invoicing.
default/* models
| Model ID | Input / 1M | Output / 1M | Use case |
|---|---|---|---|
default/gpt-oss-20b | $0.10 | $0.40 | General-purpose |
default/gpt-oss-120b | $0.35 | $1.40 | Higher-capability general model |
default/gemma-4-26b-a4b-it | $0.08 | $0.30 | Instruction-tuned |
default/gemma-4-31b | $0.12 | $0.48 | General-purpose |
default/deepseek-v4-pro | $0.45 | $1.80 | Reasoning and coding |
default/glm-5.2 | $0.30 | $1.20 | General-purpose |
default/kimi-k2.6 | $0.28 | $1.12 | Long-context general model |
default/kimi-k2.7-code | $0.32 | $1.28 | Coding |
fast/* models
| Model ID | Input / 1M | Output / 1M | Use case |
|---|---|---|---|
fast/gemma-4-31b | $0.16 | $0.64 | Lower-latency route |
fast/gpt-oss-120b | $0.45 | $1.80 | Lower-latency route |
Use a built-in model
Section titled “Use a built-in model”Set the target directly in your executor:
const upstream = "default/gpt-oss-20b";The built-in model is separate from the model ID your clients call. For example, expose assistant/support to clients and route it to default/gpt-oss-20b inside the executor.
To use OpenAI, Anthropic, Google, or another provider account of your own, see Connect providers. That bring-your-own-key setup is optional.