Models & API compatibility
Find model details, request identifiers and the intended protocol for your tool.
On this page
Current availabilityChoose a model IDImage modelsVideo modelsMatch your client to an APICheck capabilities before enabling themCompare billing separatelyCurrent availability#
The model directory distinguishes priced catalog previews from Coming soon families. No model is callable through DSSUB yet. Upstream model capabilities do not establish the capabilities of a future DSSUB route.
Choose a model ID#
| Preview request ID | Model entry | Plan points per token |
|---|---|---|
| deepseek-flash | DeepSeek V4.1 Flash | 1 |
| deepseek-v4-pro | DeepSeek V4 Pro | 3.5 |
deepseek-pro remains a legacy design alias for Pro. Prefer deepseek-v4-pro in new templates. These names do not imply an issued key has any model entitlement.
See the Flash detail page and Pro detail page for reference prices, limits awaiting verification and setup links.
Image models#
GPT Image and Seedream are planned image families. Versions, request IDs, capabilities, prices and plan inclusion are not announced.
Video models#
MiniMax Video is a separate roadmap entry from MiniMax text. Video duration, resolution, audio, request IDs and pricing require their own catalog and verification.
Match your client to an API#
| Tool | Intended API format | Intended request path |
|---|---|---|
| Codex | Responses | /v1/responses |
| Claude Code | Messages | /v1/messages |
| OpenCode, Hermes, Pi, OpenClaw, Harness | Chat Completions | /v1/chat/completions |
https://api.dssub.invalid/v1 is a non-routable template base, not a real API. A gateway must implement each format separately. See compatibility for verification scope.
Check capabilities before enabling them#
Context and maximum output, streaming, client-defined tools, reasoning fields, image input and provider-hosted tools all need verification for the exact route and client version. No DSSUB limits or reasoning levels are confirmed. Do not infer them from a model family name.
Compare billing separately#
Model IDs select models; Base, Max, Pro Max and Ultra name plan tiers. Cached input is included in input tokens and must be counted once. Compare design fees and point use with the calculator.
Guide updated November 5, 2026 · All tutorials