Skip to main content
GET /v1/models returns the live catalog: per-model pricing (the charged rate), context window, and a privacy field. The endpoint is public — fetch it without a key to browse.

​
The catalog

The launch set is curated — the models developers actually run agents on, plus the frontier set for completeness: The live set is always whatever GET /v1/models returns — it intersects this curated list with each backend’s live catalog. List from the CLI with t2 models.
Reasoning vs fast models. glm-5.2, kimi-k2.5, kimi-k2.6, and kimi-k2.7-code are reasoning models — they generate a hidden reasoning trace before answering, so they’re deeper but noticeably slower (higher latency, more tokens). They carry "reasoning": true in GET /v1/models (and a reasoning tag in t2 models). For snappy responses, reach for a non-reasoning model like gpt-oss-120b, deepseek-v4-flash, or qwen3.6-27b. We never cap max_tokens by default — the choice of depth vs speed is yours (pass your own max_tokens to bound a call).

​
Privacy tiers

Each model carries a privacy field so you can choose your posture programmatically:
  • private (default, every model) — zero data retention: providers are contractually bound not to store or train on your prompts.
  • confidential (phala/*) — inference runs in a verified GPU-TEE, with a signed receipt auto-anchored on Sui + durably stored, so you can verify it yourself any time (CLI or verify.t2000.ai). Carries a premium over private for the verifiable-privacy guarantee. See Confidential AI.

​
Pricing

One price per model, billed per token from your credit balance — the same rate shown in GET /v1/models. There’s no separate price book and no per-provider account: one key, one balance, every model. moonshotai/kimi-k2.7-code has a free daily allowance per account; for everything else, add credit (card, USDC/USDsui, or an Audric plan) at agents.t2000.ai/manage — see Authentication.