Reasoning models por dentro: o1, o3, R1, Gemini Thinking

# Claude 4.7 Opus com extended thinking
import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-opus-4-7",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 8000,   # 1k mínimo, 64k máximo
    },
    messages=[{
        "role": "user",
        "content": "Prove que sqrt(2) é irracional usando contradição.",
    }]
)

# Resposta contém blocos:
# [{"type": "thinking", "thinking": "Hmm, deixa eu pensar..."},
#  {"type": "text", "text": "Demonstração:..."}]
for block in message.content:
    if block.type == "thinking":
        print("=== THINKING ===")
        print(block.thinking)
    elif block.type == "text":
        print("=== RESPOSTA ===")
        print(block.text)

# OpenAI o3 (sem visibility do thinking)
from openai import OpenAI

client = OpenAI()

response = client.chat.completions.create(
    model="o3",
    reasoning_effort="high",     # low/medium/high
    messages=[{
        "role": "user",
        "content": "Prove que sqrt(2) é irracional usando contradição.",
    }]
)

# Não há acesso ao conteúdo do reasoning
print(response.choices[0].message.content)
print(f"Reasoning tokens: {response.usage.completion_tokens_details.reasoning_tokens}")
print(f"Output tokens: {response.usage.completion_tokens}")
# reasoning_tokens são cobrados mesmo invisíveis

# DeepSeek-R1 via API ou local
curl https://api.deepseek.com/chat/completions \
  -H "Authorization: Bearer $DEEPSEEK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-reasoner",
    "messages": [{"role": "user", "content": "Prove sqrt(2) irracional."}],
    "max_tokens": 8000
  }'

# Resposta contém "reasoning_content" + "content"
# reasoning_content tem o <think>...</think> expandido