GET
/v1/schemas/byteplus/chat/chat%2Fcompletions
Accept:
Request body
27 properties · 2 required · 13 $defs| property | type | description | constraints |
|---|---|---|---|
| model* | string | Ark model id (e.g. seed-2-0-lite-260428) or a preconfigured endpoint id. | |
| messages* | ChatCompletionMessage[] | ||
| max_tokens | integer | Maximum tokens to generate. Mutually exclusive with max_completion_tokens — sending both is a 400. | |
| temperature | number | ||
| top_p | number | ||
| stream | boolean | ||
| stop | string[] | Up to four strings that stop generation when produced. | |
| frequency_penalty | number | ||
| logit_bias | object | ||
| logprobs | boolean | ||
| top_logprobs | integer | ||
| user | string | Opaque end-user identifier forwarded for abuse monitoring. | |
| function_call | any | ||
| tools | Tool[] | ||
| tool_choice | any | ||
| stream_options | StreamOptions | ||
| presence_penalty | number | ||
| repetition_penalty | number | Penalty applied to repeated tokens. Values above 1 discourage repetition. | |
| n | integer | ||
| response_format | ResponseFormat | Structured output. Only {type: 'json_schema'} works, and only on a subset of models — Ark's own features.structured_outputs flag is unreliable (seed-2-0-lite-260428 advertises support and then 400s). Probe-verified accepting: dola-seed-2-1-turbo-260628, seed-2… | |
| parallel_tool_calls | boolean | ||
| service_tier | string | 'flex' routes to the batch queue at a lower price with no latency guarantee; 'default' is the standard online tier. | |
| thinking | Thinking | Reasoning ('deep thinking') switch. 'enabled' is the default on every model except deepseek-v3-2-251201; 'auto' is only accepted by gpt-oss-120b-250805. Cannot be combined with reasoning_effort when disabled (400 InvalidParameter). | |
| max_completion_tokens | integer | OpenAI's newer name for max_tokens; mutually exclusive with it. | |
| reasoning_effort | enum | Reasoning budget hint. 'none' and 'xhigh' are only accepted by glm-5-2-260617; 'max' by glm-5-2-260617 and the deepseek-v4-* models. Rejected in combination with thinking: {type: 'disabled'}. Absent from the Go SDK; probe-verified 2026-07-31. | none | minimal | low | medium | high | xhigh · +1 more |
| top_k | integer | Restricts sampling to the k most likely tokens. Absent from the Go SDK; probe-verified 2026-07-31. | |
| seed | integer | Best-effort determinism hint for repeated identical requests. Absent from the Go SDK; probe-verified 2026-07-31. |
Validate a payload
before you spend tokensshell
$ curl -X POST https://modelschemas.com/v1/validate \
-d '{"provider":"byteplus","endpointId":"chat/completions","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}