Updated
Reasoning effort on an abliterated LLM API
Every model here thinks before it answers, and reasoning_effort sets how much: none, low, high or max. The thinking comes back in its own field, apart from the answer.
Python
Python
https://api.silkcompute.comreply = client.chat.completions.create(
model="kimi-k3.1-pro-uncensored",
messages=[{"role": "user", "content": "Find the race condition in this function."}],
reasoning_effort="max", # "none", "low", "high" or "max"
)
message = reply.choices[0].message
print(getattr(message, "reasoning_content", None)) # the thinking
print(message.content) # the answerExample response
{
"choices": [{
"message": {
"reasoning_content": "Two writers touch count without a lock...",
"content": "The race is on count: wrap the read and write in one lock."
},
"finish_reason": "stop"
}]
}const reply = await client.chat.completions.create({
model: 'kimi-k3.1-pro-uncensored',
messages: [{ role: 'user', content: 'Find the race condition in this function.' }],
reasoning_effort: 'max', // 'none', 'low', 'high' or 'max'
})
const message = reply.choices[0].message
console.log(message.reasoning_content) // the thinking
console.log(message.content) // the answerExample response
{
"choices": [{
"message": {
"reasoning_content": "Two writers touch count without a lock...",
"content": "The race is on count: wrap the read and write in one lock."
},
"finish_reason": "stop"
}]
}curl https://api.silkcompute.com/chat/completions \
-H "Authorization: Bearer $SILK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "kimi-k3.1-pro-uncensored",
"reasoning_effort": "max",
"messages": [
{"role": "user", "content": "Find the race condition in this function."}
]
}'Example response
{
"choices": [{
"message": {
"reasoning_content": "Two writers touch count without a lock...",
"content": "The race is on count: wrap the read and write in one lock."
},
"finish_reason": "stop"
}]
}response = client.responses.create(
model="deepseek-v4.1-flash-1017-uncensored",
input="Find the race condition in this function.",
reasoning={"effort": "high"},
store=False,
)message = client.messages.create(
model="deepseek-v4.1-flash-1017-uncensored",
max_tokens=16000,
thinking={"type": "enabled", "budget_tokens": 8192},
output_config={"effort": "high"},
messages=[{"role": "user", "content": "Find the race condition in this function."}],
)How it works
Levelsnone, low, high, maxOther names map to the nearest level, so
minimal runs as low, medium as high and xhigh as max.max_tokensThinking countsThinking counts toward
max_tokens. If it runs out, the answer is empty and finish_reason is length.FormatsSet it in any formatEach format has its own field, listed below.
Per format
| Format | Set it with | Thinking comes back in |
|---|---|---|
| Chat Completions | reasoning_effort | message.reasoning_content |
| Responses | reasoning.effort | A reasoning output item |
| Messages | output_config.effort, with thinking on | thinking blocks |
Default effort per model
| Model | Default effort |
|---|---|
| Kimi K3.1 Pro | max |
| DeepSeek V4.1 Flash 1017 | high |
| DeepSeek V4.1 Flash 1017 Fast | high |
| GLM-5.3 | high |
| GLM-5.3 Flash | high |