Reasoning models

Reasoning models perform internal work before answering difficult math, coding, and logic tasks. PleumRouter automatically adapts reasoning requests to OpenAI and Anthropic models, and passes values such as signatures and encrypted data through without interpreting them.

How it works#

OpenAI reasoning models such as gpt-5.2, gpt-5.4, gpt-5.5, gpt-5.6-sol, and gpt-6-sol are sent without temperature, with max_tokens rewritten to max_completion_tokens. That rewrite applies only to OpenAI reasoning models.

gpt-5.4, gpt-5.4-mini, and gpt-5.5 are forced-reasoning models: none is removed, while low/medium/highis forwarded. gpt-5.2 and GPT-5.6 (sol/terra/luna) are optional and accept minimal…max. GPT-5.6 also accepts reasoning_mode (standard|pro) and service_tier.

Claude reasoning behavior differs by model. A model whose reasoning is always on cannot be switched off with none. On other models, none becomes a request to turn reasoning off, and low/medium/high become a reasoning level suited to the model. Models that take their reasoning budget as an explicit token count use 20%/50%/80% of max_tokens (minimum 1,024, maximum 32,000, always below max_tokens) and reject a max_tokens of 1,024 or less with 400. Check each model's support in the supports_thinking, thinking_forced, and supported_parameters fields of the GET /v1/models response.

{
  "model": "gpt-5.6-sol",
  "messages": [
    {"role": "user", "content": "Prove that there are infinitely many primes."}
  ],
  "max_completion_tokens": 4096,
  "reasoning_effort": "xhigh",
  "reasoning_mode": "pro"
}

Request#

curl https://apirouter.pleum.ai/v1/chat/completions \
  -H "Authorization: Bearer $PLEUM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-sol",
    "messages": [
      {"role": "user", "content": "Prove that there are infinitely many primes."}
    ],
    "max_tokens": 4096,
    "reasoning_effort": "xhigh",
    "reasoning_mode": "pro"
  }'