Skip to main content
Reasoning fine-tuning adapts a model that supports chain-of-thought reasoning. By providing reasoning or reasoning_content alongside the final assistant response, you shape how the model thinks through problems before producing an answer. This page covers the reasoning data shape, supported models, and launch parameters.
Reasoning models should always be fine-tuned with reasoning data. Training a reasoning model without it can degrade its reasoning ability. If your dataset doesn’t include reasoning, use an instruct model instead.

​
Supported models

The following models support reasoning fine-tuning. See supported models for context lengths and batch limits.
OrganizationModelAPI ID
NVIDIANVIDIA Nemotron 3 Nano Omni 30B A3B Reasoning BF16nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16
QwenQwen3.5 397B A17BQwen/Qwen3.5-397B-A17B
QwenQwen3.5 122B A10BQwen/Qwen3.5-122B-A10B
QwenQwen3.5 35B A3BQwen/Qwen3.5-35B-A3B
QwenQwen3.5 35B A3B BaseQwen/Qwen3.5-35B-A3B-Base
QwenQwen3.5 27BQwen/Qwen3.5-27B
QwenQwen3.5 9BQwen/Qwen3.5-9B
QwenQwen3.5 4BQwen/Qwen3.5-4B
QwenQwen3.5 2BQwen/Qwen3.5-2B
QwenQwen3.5 0.8BQwen/Qwen3.5-0.8B
QwenQwen3.6 35B A3BQwen/Qwen3.6-35B-A3B
QwenQwen3 Next 80B A3B ThinkingQwen/Qwen3-Next-80B-A3B-Thinking
QwenQwen3 0.6BQwen/Qwen3-0.6B
QwenQwen3 0.6B BaseQwen/Qwen3-0.6B-Base
QwenQwen3 1.7BQwen/Qwen3-1.7B
QwenQwen3 1.7B BaseQwen/Qwen3-1.7B-Base
QwenQwen3 4BQwen/Qwen3-4B
QwenQwen3 4B BaseQwen/Qwen3-4B-Base
QwenQwen3 8BQwen/Qwen3-8B
QwenQwen3 8B BaseQwen/Qwen3-8B-Base
QwenQwen3 14BQwen/Qwen3-14B
QwenQwen3 14B BaseQwen/Qwen3-14B-Base
QwenQwen3 32BQwen/Qwen3-32B
QwenQwen3 30B A3B BaseQwen/Qwen3-30B-A3B-Base
QwenQwen3 30B A3BQwen/Qwen3-30B-A3B
QwenQwen3 235B A22BQwen/Qwen3-235B-A22B
Z.aiGLM 5.1zai-org/GLM-5.1
Z.aiGLM 5zai-org/GLM-5
Z.aiGLM 4.7zai-org/GLM-4.7
Z.aiGLM 4.6zai-org/GLM-4.6
OpenAIGPT-OSS 20Bopenai/gpt-oss-20b
OpenAIGPT-OSS 120Bopenai/gpt-oss-120b
GoogleGemma 4 31B ITgoogle/gemma-4-31B-it
GoogleGemma 4 31B IT VLMgoogle/gemma-4-31B-it-VLM
GoogleGemma 4 26B A4B ITgoogle/gemma-4-26B-A4B-it

​
Prepare your data

Prepare data in a JSONL file. Each assistant message should carry the chain of thought in a reasoning (or reasoning_content) field and the final answer in content.

​
Conversational format

{
  "messages": [
    {"role": "user", "content": "What is the capital of France?"},
    {
      "role": "assistant",
      "reasoning": "The user is asking about the capital of France. France is a country in Western Europe. Its capital city is Paris, which has been the capital since the 10th century.",
      "content": "The capital of France is Paris."
    }
  ]
}
When fine-tuning reasoning models on conversational data, only the last assistant message is trained on by default. For multi-turn reasoning, split the conversation so each assistant message is the final message in its own example.

​
Preference format

For preference fine-tuning, both outputs carry reasoning. See preference tuning for the broader DPO workflow. 
{
  "input": {
    "messages": [
      {"role": "user", "content": "What is the capital of France?"}
    ]
  },
  "preferred_output": [
    {
      "role": "assistant",
      "reasoning": "France is in Western Europe. Its capital is Paris.",
      "content": "The capital of France is Paris."
    }
  ],
  "non_preferred_output": [
    {
      "role": "assistant",
      "reasoning": "Let me think about European capitals.",
      "content": "The capital of France is Berlin."
    }
  ]
}

​
Validate and upload

Upload your data using the Together Python/TypeScript SDK or the Together CLI: 
from together import Together

client = Together()

train_file = client.files.upload(
    file="reasoning_dataset.jsonl",
    purpose="fine-tune",
    check=True,
)
print(train_file.id)

import Together from "together-ai";
import fs from "node:fs";

const client = new Together();

const trainFile = await client.files.upload({
  file: fs.createReadStream("reasoning_dataset.jsonl"),
  purpose: "fine-tune",
});
console.log(trainFile.id);

tg files check "reasoning_dataset.jsonl"
tg files upload "reasoning_dataset.jsonl"

​
Launch the job

LoRA is the default. Pass lora=False for full fine-tuning. 
job = client.fine_tuning.create(
    training_file=train_file.id,
    model="Qwen/Qwen3-8B",
    lora=True,
)
print(job.id)

const job = await client.fineTuning.create({
  training_file: trainFile.id,
  model: "Qwen/Qwen3-8B",
  lora: true,
});
console.log(job.id);

tg fine-tuning create \
  --training-file "<FILE_ID>" \
  --model "Qwen/Qwen3-8B" \
  --lora
For details on every available parameter, see the API reference.

​
Watch and deploy

Reasoning jobs use the same lifecycle as text jobs:
  • Poll the job with the SDK or CLI. Expect 10 to 30 minutes for a LoRA job on an 8B model with a few thousand examples.
  • Deploy the result on a dedicated endpoint.
  • Call the endpoint with the same chat-completions shape. The model emits reasoning_content alongside content for clients that surface it. See Inference → Reasoning for details.