MiniMax

Runtime provider id: minimax. Served through the AI SDK Anthropic adapter (@ai-sdk/anthropic via the Strands VercelModel), because it reads token usage from the stream’s final message_delta — MiniMax reports input_tokens: 0 in message_start, which the native Anthropic model would trust.

Provider options

Field	Type	Notes
apiKey	string	Required (or set via environment).
baseURL	string	Optional. Defaults to the Anthropic-compatible endpoint https://api.minimax.io/anthropic; override to reach MiniMax through a gateway (e.g. LiteLLM).
headers	Record<string, string>	Optional. Extra HTTP headers.
reasoning	object	Optional. See Reasoning.

Reasoning

Providing reasoning enables MiniMax’s adaptive thinking (thinking: { type: "adaptive" } with output_config.effort); omit reasoning to leave thinking at the model default. effort defaults to medium (minimal maps to low); display is forwarded when set (MiniMax accepts display, unlike the native Anthropic API).

Example configs

{
  "name": "MiniMax",
  "provider": "minimax",
  "options": {
    "apiKey": "..."
  }
}

{
  "name": "MiniMax M3",
  "provider": "MiniMax",
  "options": {
    "model": "MiniMax-M3",
    "temperature": 1,
    "maxTokens": 4096
  },
  "default": true
}

Through a LiteLLM gateway, with reasoning display:

{
  "name": "MiniMax via LiteLLM",
  "provider": "minimax",
  "options": {
    "apiKey": "...",
    "baseURL": "https://litellm.internal/anthropic",
    "reasoning": { "effort": "high", "display": "summarized" }
  }
}

{
  "name": "MiniMax M3 (thinking)",
  "provider": "MiniMax via LiteLLM",
  "options": {
    "model": "MiniMax-M3"
  }
}