supported_endpoint_types is a discovery signal, not an end-to-end tool guarantee. Choose models by endpoint, streaming behavior, tool calling, and output structure.
Endpoint selection
| Endpoint | Use it when | Typical tools |
|---|---|---|
POST /v1/chat/completions | OpenAI-compatible providers, text chat, default coding-agent path | opencode OpenAI-compatible, Kilo OpenAI Compatible, Hermes custom endpoint, OpenClaw openai-completions |
POST /v1/messages | Claude/Anthropic providers and Claude Messages request shape | Claude Code, Kilo Anthropic, opencode Anthropic, OpenClaw anthropic-messages |
POST /v1/responses | A tool or SDK explicitly calls OpenAI Responses API, such as responses.create | OpenAI Responses SDK, Agents SDK, Responses runtime |
Model matrix
| Model | Chat Completions | Anthropic Messages | Responses | Recommendation |
|---|---|---|---|---|
glm-4.5-air | Available | Recommended | Not supported | Use with OpenAI-compatible or Anthropic tools |
glm-4.7 | Available | Recommended | Not supported | Use with OpenAI-compatible or Anthropic tools |
glm-4.7-flash | Available | Recommended | Not supported | Good for lightweight tasks |
glm-5 | Available | Recommended | Not supported | Good for reasoning tasks |
glm-5-turbo | Available | Recommended | Not supported | Good lightweight model |
glm-5.2 | Available | Recommended | Not supported | Recommended Anthropic-tool default |
MiniMax-M2.5 | Available | Recommended | Available | Increase max_output_tokens for Responses |
MiniMax-M2.5-highspeed | Available | Recommended | Available | Increase max_output_tokens for Responses |
MiniMax-M2.7 | Available | Recommended | Available | Increase max_output_tokens for Responses |
MiniMax-M2.7-highspeed | Available | Recommended | Available | Increase max_output_tokens for Responses |
MiniMax-M3 | Available | Recommended | Available | Recommended MiniMax tool model |
qwen3-coder-plus | Available | Recommended | Available | Good for coding and tool workflows |
qwen3-max | Available | Recommended | Available | Good for general reasoning and review |
qwen3.5-flash | Available | Recommended | Available | Responses may include reasoning summary |
qwen3.6-plus | Available | Recommended | Available | Responses may include reasoning summary |
deepseek-v3-2-251201 | Available | Recommended | Available | Passed opencode Anthropic-provider tool-workflow validation |
seed-1-6-250915 | Available | Recommended | Available | Use with OpenAI-compatible, Anthropic, or Responses |
seed-1-6-flash-250715 | Available | Not supported | Available | The upstream provider does not support /v1/messages; prefer Chat Completions or Responses |
seed-1-8-251228 | Available | Recommended | Available | Use with OpenAI-compatible, Anthropic, or Responses |
seed-2-0-lite-260228 | Available | Recommended | Available | Use with OpenAI-compatible, Anthropic, or Responses |
seed-2-0-mini-260215 | Available | Recommended | Available | Use with OpenAI-compatible, Anthropic, or Responses |
seed-2-0-pro-260328 | Available | Recommended | Available | Use with OpenAI-compatible, Anthropic, or Responses |
Ecosystem tool guidance
| Tool | Preferred endpoint | Suggested models |
|---|---|---|
| Claude Code | /v1/messages | glm-5.2, MiniMax-M3 |
| opencode | /v1/chat/completions | glm-5.2, MiniMax-M3 |
| Kilo Code | /v1/chat/completions or /v1/messages | glm-5.2, MiniMax-M3 |
| Hermes Agent | /v1/chat/completions | glm-5.2, MiniMax-M3, seed-2-0-pro-260328 |
| OpenClaw | /v1/chat/completions or /v1/messages | glm-5.2, MiniMax-M3 |
| OpenAI Responses SDK / Agents SDK | /v1/responses | Qwen, MiniMax, Seed, deepseek-v3-2-251201 |
Known boundaries
- GLM models are not currently Responses-path models.
seed-1-6-flash-250715is not supported by its upstream provider on/v1/messages; other Seed models passed opencode Anthropic-provider validation.deepseek-v3-2-251201passed opencode Anthropic-provider tool-workflow validation.- In Responses, some MiniMax, Qwen, and Seed models return reasoning or reasoning summary. Extract user-visible text from
output[].