Models

Sable exposes a curated catalog under stable identifiers so model upgrades don't break your agents. List them at runtime with GET /v1/models.

Each entry carries pricing (prompt_usd_per_mtok and completion_usd_per_mtok, in USD per 1M tokens) and an open_weight flag. See Cost & metering.

Open-weight vs. vendor-hosted

/v1/models marks every model open_weight: true or false. It matters for the privacy roadmap:

• open_weight: true means the weights are public (Llama, DeepSeek, Mistral, Qwen, Nomic), so the model can run inside a Sable-verified enclave. One does today — sable-confidential-24b runs in an attested Intel TDX enclave on the confidential tier. The rest are served on the standard path for now.
• open_weight: false means the model only runs on its vendor's servers (OpenAI's gpt-4o-mini and the text-embedding-3 models). OpenAI sees its input, and it will never be confidential beyond encrypt-in-transit. We'd rather tell you that than ship a closed model under a privacy banner, so these stay standard tier only.

The thing to know: on standard tier every model runs on a third-party host that decrypts the prompt first (see the privacy contract). The confidential tier is the exception — today that's sable-confidential-24b, which runs in an attested enclave the host can't see into, proven in the signed receipt. The open_weight flag tells you which models can follow it there.

Chat

| Sable id | Family | Context | Open weight | Default tier | | --- | --- | --- | --- | --- | | sable-confidential-24b | Mistral | 32k | ✅ | confidential (attested TDX, live) | | sable-llama-3.3-70b | Meta Llama | 131k | ✅ | standard | | sable-llama-3.1-8b | Meta Llama | 131k | ✅ | standard | | sable-deepseek-v3 | DeepSeek | 64k | ✅ | standard | | sable-mistral-large | Mistral | 128k | ✅ | standard | | sable-qwen-2.5-72b | Qwen | 32k | ✅ | standard | | sable-gpt-4o-mini | OpenAI | 128k | ❌ OpenAI-hosted | standard |

Embeddings

| Sable id | Family | Max input | Open weight | Default tier | | --- | --- | --- | --- | --- | | sable-embed-3-small | OpenAI | 8k tokens | ❌ OpenAI-hosted | standard | | sable-embed-3-large | OpenAI | 8k tokens | ❌ OpenAI-hosted | standard | | sable-embed-nomic | Nomic | 8k tokens | ✅ | standard |