Models that think before answering
Reasoning
Use reasoning-capable models, control effort and token budgets, and stream the model's reasoning as it works.
reasoning_effort: "medium"
"reasoning": "Common denominator is 12…2/3 = 8/12, 1/4 = 3/12, 5/6 = 10/128 + 3 + 10 = 21 → 21/12 = 7/4""content": "The sum is 7/4 (1.75).""usage": { "reasoning_tokens": 182 }
Same parameter on every provider
01
Effort control
Low, medium or high.
02
Token budgets
Cap thinking where supported.
03
Streamed thoughts
Watch reasoning arrive.
04
Usage tracking
Reasoning tokens reported.
Built for
Math & logic
Step-by-step problem solving
Code review
Deeper analysis on hard diffs
Planning agents
Think before acting
Set the effort
terminal
1curl https://api.llmgateway.io/v1/chat/completions \2 -H "Authorization: Bearer $LLM_GATEWAY_API_KEY" \3 -H "Content-Type: application/json" \4 -d '{5 "model": "gpt-oss-120b",6 "messages": [7 {8 "role": "user",9 "content": "What is 2/3 + 1/4 + 5/6?"10 }11 ],12 "reasoning_effort": "medium"13 }'