OpenAI Chat Completions

Field-for-field compatible with OpenAI /v1/chat/completions, including streaming, tool calls, reasoning parameters and structured output.

Endpoint

POST/v1/chat/completions

Field	Type		Description
model	string	required	Model name or alias, see the model catalogue.
messages	array	required	Conversation messages, role ∈ system / user / assistant / tool; content supports text and images (image_url).
stream	boolean	optional	When true, returns SSE chunks as data: {...}; the last frame carries usage, followed by data: [DONE].
stream_options.include_usage	boolean	optional	Attach usage to the final frame when streaming.
max_completion_tokens	integer	optional	Output token cap (max_tokens is accepted too).
temperature / top_p	number	optional	Sampling parameters, passed through upstream.
tools / tool_choice	array / string	optional	Function-calling definitions and policy, passed through upstream.
response_format	object	optional	json_object or json_schema structured output (where the model supports it).
reasoning_effort	string	optional	Reasoning effort low / medium / high (reasoning models).

Requestbash

curl https://www.link42.ai/v1/chat/completions \
  -H "Authorization: Bearer $LINK42_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "stream": true,
    "stream_options": {"include_usage": true},
    "messages": [
      {"role": "system", "content": "You are a concise assistant."},
      {"role": "user", "content": "List three benefits of using an API gateway"}
    ]
  }'

Response (non-streaming)json

{
  "id": "chatcmpl-8f1c…",
  "object": "chat.completion",
  "created": 1788300000,
  "model": "glm-5.2",
  "choices": [{"index": 0, "message": {"role": "assistant", "content": "…"}, "finish_reason": "stop"}],
  "usage": {"prompt_tokens": 38, "completion_tokens": 120, "total_tokens": 158}
}

The token counts in usage are the billing basis; cached input is charged at the cache_read price.
If the client drops the connection mid-stream, the part already generated is still settled at actual usage.

Model list

GET/v1/models

Returns the models available to the key's group in the OpenAI shape {object:"list", data:[{id, object:"model", owned_by}]}.

Model reference

The first column is the catalogue id you put in `model`. Deployment-configured aliases such as seed-2.0-pro also work; the model plaza shows the ones your deployment defines.

This table includes migration references and models not yet activated; it does not promise every row is callable. /anthropic/v1/messages and /v1/responses support also depends on the selected model and your key group's current capabilities. Check GET /v1/models and test the endpoint before use. Unspecified fields follow the endpoint reference.

Model id	Name	Upstream	Billing	Fields that differ	Limits and gotchas
seed-2-0-pro-260328	Dola-Seed-2.0-pro	BytePlus Ark	Per token	messages[].content accepts text, image_url and input_audio; the reply message carries reasoning_content; usage carries prompt_tokens_details.cached_tokens / audio_tokens and completion_tokens_details.reasoning_tokens	Above 128,000 input tokens in one request the whole call is priced at double rate; image input counts inside prompt_tokens at the text input price; audio input tokens are counted twice under the legacy_v1 profile
seed-2-0-lite-260428	Dola-Seed-2.0-lite	BytePlus Ark	Per token	Same as 2.0-pro but without image input; identical usage detail	Same long-context threshold and legacy_v1 audio double count as 2.0-pro
seed-2-0-mini-260428	Dola-Seed-2.0-mini	BytePlus Ark	Per token	Same as 2.0-lite; accepts input_audio audio input	Same long-context threshold and legacy_v1 audio double count as 2.0-pro
dola-seed-2-1-turbo-260628	Dola-Seed-2.1-turbo	BytePlus Ark	Per token	Plain chat fields, no extensions; usage carries only prompt / completion plus cached_tokens and reasoning_tokens detail	No long-context surcharge tier
deepseek-v4-pro-260425	DeepSeek-V4-Pro	BytePlus Ark	Per token	Plain chat fields, no extensions	No long-context surcharge tier
deepseek-v4-flash-260425	DeepSeek-V4-Flash	BytePlus Ark	Per token	Plain chat fields, no extensions	No long-context surcharge tier
glm-5-2-260617	GLM-5.2	BytePlus Ark	Per token	Plain chat fields, no extensions	No long-context surcharge tier
deepseek-v4-1-flash	DeepSeek-V4.1-Flash	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; text, vision and tools; cache hits have their own rate	Upstream deepseek-v4-1-flash-260910; check the live catalogue and routing state for availability, not this document. Weekday peak pricing is 09:00–12:00 and 14:00–18:00 Beijing time.
deepseek-v4-pro	DeepSeek-V4-Pro GA	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; cache hits are priced separately	Upstream deepseek-v4-pro-ga-260813; check the live catalogue and routing state for availability, not this document.
deepseek-v4-flash	DeepSeek-V4-Flash GA	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; cache hits are priced separately	Currently maps upstream to deepseek-v4-flash-ga-260731; see the live marketplace for the USD rate.
glm-5-3-flash	GLM-5.3-Flash	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; cache hits are priced separately	Upstream glm-5-3-flash-260828; check the live catalogue and routing state for availability, not this document.
glm-5.2	Alibaba GLM-5.2	Alibaba Bailian (OpenAI-compatible endpoint)	Per token	Same single-turn prompt shape on the private protocol as Alibaba DeepSeek-V4-Pro	cached_tokens are billed at the cache-read price; no long-context surcharge tier

The long-context surcharge applies to the whole call: once one request exceeds the threshold, that call's input and output are both settled at the raised rate — the surcharge is not limited to the tokens above the threshold.
Counting audio input tokens twice is historical behaviour kept only in the legacy_v1 profile (migrated ak- keys). Newly created sk- keys are standard_v1 and count them once.
The unsuffixed deepseek-v4-pro and deepseek-v4-flash models are bound to domestic Volcengine Ark; the dated -260425 rows are separate historical BytePlus entries. Parameter documentation does not imply current activation. Check the live marketplace and your API key's group catalogue.

Endpoint

POST/v1/chat/completions

Field	Type		Description
model	string	required	Model name or alias, see the model catalogue.
messages	array	required	Conversation messages, role ∈ system / user / assistant / tool; content supports text and images (image_url).
stream	boolean	optional	When true, returns SSE chunks as data: {...}; the last frame carries usage, followed by data: [DONE].
stream_options.include_usage	boolean	optional	Attach usage to the final frame when streaming.
max_completion_tokens	integer	optional	Output token cap (max_tokens is accepted too).
temperature / top_p	number	optional	Sampling parameters, passed through upstream.
tools / tool_choice	array / string	optional	Function-calling definitions and policy, passed through upstream.
response_format	object	optional	json_object or json_schema structured output (where the model supports it).
reasoning_effort	string	optional	Reasoning effort low / medium / high (reasoning models).

Requestbash

curl https://www.link42.ai/v1/chat/completions \
  -H "Authorization: Bearer $LINK42_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.2",
    "stream": true,
    "stream_options": {"include_usage": true},
    "messages": [
      {"role": "system", "content": "You are a concise assistant."},
      {"role": "user", "content": "List three benefits of using an API gateway"}
    ]
  }'

Response (non-streaming)json

{
  "id": "chatcmpl-8f1c…",
  "object": "chat.completion",
  "created": 1788300000,
  "model": "glm-5.2",
  "choices": [{"index": 0, "message": {"role": "assistant", "content": "…"}, "finish_reason": "stop"}],
  "usage": {"prompt_tokens": 38, "completion_tokens": 120, "total_tokens": 158}
}

The token counts in usage are the billing basis; cached input is charged at the cache_read price.

If the client drops the connection mid-stream, the part already generated is still settled at actual usage.

Model reference

The first column is the catalogue id you put in `model`. Deployment-configured aliases such as seed-2.0-pro also work; the model plaza shows the ones your deployment defines.

Model id	Name	Upstream	Billing	Fields that differ	Limits and gotchas
seed-2-0-pro-260328	Dola-Seed-2.0-pro	BytePlus Ark	Per token	messages[].content accepts text, image_url and input_audio; the reply message carries reasoning_content; usage carries prompt_tokens_details.cached_tokens / audio_tokens and completion_tokens_details.reasoning_tokens	Above 128,000 input tokens in one request the whole call is priced at double rate; image input counts inside prompt_tokens at the text input price; audio input tokens are counted twice under the legacy_v1 profile
seed-2-0-lite-260428	Dola-Seed-2.0-lite	BytePlus Ark	Per token	Same as 2.0-pro but without image input; identical usage detail	Same long-context threshold and legacy_v1 audio double count as 2.0-pro
seed-2-0-mini-260428	Dola-Seed-2.0-mini	BytePlus Ark	Per token	Same as 2.0-lite; accepts input_audio audio input	Same long-context threshold and legacy_v1 audio double count as 2.0-pro
dola-seed-2-1-turbo-260628	Dola-Seed-2.1-turbo	BytePlus Ark	Per token	Plain chat fields, no extensions; usage carries only prompt / completion plus cached_tokens and reasoning_tokens detail	No long-context surcharge tier
deepseek-v4-pro-260425	DeepSeek-V4-Pro	BytePlus Ark	Per token	Plain chat fields, no extensions	No long-context surcharge tier
deepseek-v4-flash-260425	DeepSeek-V4-Flash	BytePlus Ark	Per token	Plain chat fields, no extensions	No long-context surcharge tier
glm-5-2-260617	GLM-5.2	BytePlus Ark	Per token	Plain chat fields, no extensions	No long-context surcharge tier
deepseek-v4-1-flash	DeepSeek-V4.1-Flash	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; text, vision and tools; cache hits have their own rate	Upstream deepseek-v4-1-flash-260910; check the live catalogue and routing state for availability, not this document. Weekday peak pricing is 09:00–12:00 and 14:00–18:00 Beijing time.
deepseek-v4-pro	DeepSeek-V4-Pro GA	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; cache hits are priced separately	Upstream deepseek-v4-pro-ga-260813; check the live catalogue and routing state for availability, not this document.
deepseek-v4-flash	DeepSeek-V4-Flash GA	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; cache hits are priced separately	Currently maps upstream to deepseek-v4-flash-ga-260731; see the live marketplace for the USD rate.
glm-5-3-flash	GLM-5.3-Flash	Volcengine Ark, Beijing	Native CNY list price converted to USD using today's checked SAFE quote, then the effective discount (user if set, otherwise model)	Standard chat / Responses; cache hits are priced separately	Upstream glm-5-3-flash-260828; check the live catalogue and routing state for availability, not this document.
glm-5.2	Alibaba GLM-5.2	Alibaba Bailian (OpenAI-compatible endpoint)	Per token	Same single-turn prompt shape on the private protocol as Alibaba DeepSeek-V4-Pro	cached_tokens are billed at the cache-read price; no long-context surcharge tier

The long-context surcharge applies to the whole call: once one request exceeds the threshold, that call's input and output are both settled at the raised rate — the surcharge is not limited to the tokens above the threshold.

Counting audio input tokens twice is historical behaviour kept only in the legacy_v1 profile (migrated ak- keys). Newly created sk- keys are standard_v1 and count them once.

The unsuffixed deepseek-v4-pro and deepseek-v4-flash models are bound to domestic Volcengine Ark; the dated -260425 rows are separate historical BytePlus entries. Parameter documentation does not imply current activation. Check the live marketplace and your API key's group catalogue.