Enable extended thinking mode to see the model's reasoning process for complex problems. Better transparency, better results.
Extended thinking allows the model to "think out loud" before generating a response. This intermediate reasoning is visible to you and often leads to more accurate answers.
from mythicdot import MythicDot
client = MythicDot()
# Enable extended thinking
response = client.messages.create(
model="mythic-4",
max_tokens=16000,
thinking={
"type": "enabled",
"budget_tokens": 10000 # Max tokens for thinking
},
messages=[
{
"role": "user",
"content": "Solve this complex math problem..."
}
]
)
# Access thinking and response separately
for block in response.content:
if block.type == "thinking":
print("Thinking:", block.thinking)
elif block.type == "text":
print("Response:", block.text)
Complex calculations, word problems, and logical reasoning benefit from step-by-step thinking.
Working through code logic to identify bugs and understand program flow.
Multi-step problems that require breaking down into smaller parts.
Data analysis and decision-making where showing work is important.
Extended thinking uses additional tokens for the reasoning process. Set a budget_tokens limit to control costs. Thinking tokens are charged at the same rate as output tokens. This feature works best with our most capable models (Mythic-4 and above).
Enable extended thinking for your complex reasoning tasks.