---
license: mit
task_categories:
  - text-generation
language:
  - en
tags:
  - reasoning
  - sft
  - distillation
  - merged
  - chain-of-thought
size_categories:
  - 1M<n<10M
---

# Mega Merged Reasoning SFT Dataset

Auto-merged from **44 source datasets** → **3,681,880 rows**.

## Schema

Every row is:

```json
{
  "messages": [
    {"role": "system",    "content": "..."},
    {"role": "user",      "content": "..."},
    {"role": "assistant", "content": "<think>\n...\n</think>\n\nFinal answer..."}
  ],
  "thinking": "extracted chain-of-thought (or null)",
  "source":   "org/dataset-name"
}
```

- **`messages`** — standard OpenAI / TRL chat format
- **`thinking`** — isolated CoT when available (also embedded inline in `<think>` tags)
- **`source`** — which original repo this row came from

## Formats auto-detected

| # | Strategy | Description |
|---|----------|-------------|
| 1 | `messages` list | OpenAI-style `role`/`content` dicts |
| 2 | `conversations` list | ShareGPT-style `from`/`value` dicts |
| 3 | `conversation` list | Variant key name |
| 4 | `chat`/`turns`/`dialog` list | Other conversation keys |
| 5 | `messages` with `reasoning_content` | OpenMed / R1 style per-turn reasoning |
| 6 | `prompt` + `response` | Flat pair |
| 7 | `instruction` + `output` | Alpaca-style flat pair |
| 8 | `question` + `answer` | QA-style flat pair |
| 9 | `query` + `completion` | Completion-style flat pair |
| 10 | `problem` + `solution` | Math-style flat pair |
| 11 | `input` + `output` | Generic flat pair |
| 12 | Any flat pair + `thinking`/`cot`/`reasoning` | Separate CoT column |
| 13 | `complex_cot` + `response` | Medical-o1 style |
| 14 | Nested `provenance.prompt` | Crownelius coding datasets |
| 15 | `system` / `system_prompt` field | System message extraction |
| 16 | ChatML raw text | `<|im_start|>` parsing |
| 17 | Alpaca raw text | `### Instruction:` parsing |
| 18 | Inline `<think>...</think>` | Extracted + preserved |
| 19 | JSON-string conversation | Stringified list parsing |
| 20 | Single `text` blob | Last-resort extraction |
| 21 | Brute-force string scan | Two longest strings → user/assistant |
| 22 | Prompt-only (no response) | Kept as user-only row |

## Source datasets

- [Roman1111111/claude-opus-4.6-10000x](https://huggingface.co/datasets/Roman1111111/claude-opus-4.6-10000x)
- [nohurry/Opus-4.6-Reasoning-3000x-filtered](https://huggingface.co/datasets/nohurry/Opus-4.6-Reasoning-3000x-filtered)
- [Jackrong/Qwen3.5-reasoning-700x](https://huggingface.co/datasets/Jackrong/Qwen3.5-reasoning-700x)
- [Roman1111111/gpt-5.4-step-by-step-reasoning](https://huggingface.co/datasets/Roman1111111/gpt-5.4-step-by-step-reasoning)
- [Roman1111111/gemini-3.1-pro-hard-high-reasoning](https://huggingface.co/datasets/Roman1111111/gemini-3.1-pro-hard-high-reasoning)
- [Roman1111111/gemini-3-pro-10000x-hard-high-reasoning](https://huggingface.co/datasets/Roman1111111/gemini-3-pro-10000x-hard-high-reasoning)
- [TeichAI/claude-4.5-opus-high-reasoning-250x](https://huggingface.co/datasets/TeichAI/claude-4.5-opus-high-reasoning-250x)
- [TeichAI/glm-4.6-250x](https://huggingface.co/datasets/TeichAI/glm-4.6-250x)
- [TeichAI/claude-sonnet-4.5-high-reasoning-250x](https://huggingface.co/datasets/TeichAI/claude-sonnet-4.5-high-reasoning-250x)
- [TeichAI/gpt-5-codex-250x](https://huggingface.co/datasets/TeichAI/gpt-5-codex-250x)
- [TeichAI/gemini-2.5-flash-lite-2509-preview-1000x](https://huggingface.co/datasets/TeichAI/gemini-2.5-flash-lite-2509-preview-1000x)
- [TeichAI/gpt-5-codex-1000x](https://huggingface.co/datasets/TeichAI/gpt-5-codex-1000x)
- [TeichAI/kimi-k2-thinking-1000x](https://huggingface.co/datasets/TeichAI/kimi-k2-thinking-1000x)
- [mfielding92/glm-4.7-multiturn-cot-jackrong-openai-format](https://huggingface.co/datasets/mfielding92/glm-4.7-multiturn-cot-jackrong-openai-format)
- [mfielding92/medical-o1-reasoning-SFT-jsonl](https://huggingface.co/datasets/mfielding92/medical-o1-reasoning-SFT-jsonl)
- [mfielding92/claude-3.7-sonnet-reasoning](https://huggingface.co/datasets/mfielding92/claude-3.7-sonnet-reasoning)
- [mfielding92/claude-3.7-sonnet-reasoning-qa](https://huggingface.co/datasets/mfielding92/claude-3.7-sonnet-reasoning-qa)
- [mfielding92/glm-4.7-2048-reasoning-1000x](https://huggingface.co/datasets/mfielding92/glm-4.7-2048-reasoning-1000x)
- [mfielding92/gemini-3.1-pro-2048-reasoning-1100x](https://huggingface.co/datasets/mfielding92/gemini-3.1-pro-2048-reasoning-1100x)
- [TeichAI/sherlock-think-alpha-1000x](https://huggingface.co/datasets/TeichAI/sherlock-think-alpha-1000x)
- [TeichAI/gemini-2.5-flash-11000x](https://huggingface.co/datasets/TeichAI/gemini-2.5-flash-11000x)
- [TeichAI/gpt-5.1-high-reasoning-1000x](https://huggingface.co/datasets/TeichAI/gpt-5.1-high-reasoning-1000x)
- [TeichAI/deepseek-v3.2-speciale-openr1-math-3k](https://huggingface.co/datasets/TeichAI/deepseek-v3.2-speciale-openr1-math-3k)
- [TeichAI/deepseek-v3.2-speciale-1000x](https://huggingface.co/datasets/TeichAI/deepseek-v3.2-speciale-1000x)
- [TeichAI/deepseek-v3.2-speciale-OpenCodeReasoning-3k](https://huggingface.co/datasets/TeichAI/deepseek-v3.2-speciale-OpenCodeReasoning-3k)
- [Crownelius/High-Coder-Reasoning-Multi-Turn](https://huggingface.co/datasets/Crownelius/High-Coder-Reasoning-Multi-Turn)
- [Crownelius/GLM-5.0-25000x](https://huggingface.co/datasets/Crownelius/GLM-5.0-25000x)
- [Crownelius/Opus-4.5-3000x](https://huggingface.co/datasets/Crownelius/Opus-4.5-3000x)
- [Crownelius/Creative-Writing-High-Quality-1300x](https://huggingface.co/datasets/Crownelius/Creative-Writing-High-Quality-1300x)
- [Crownelius/Creative-Writing-Gemini3Pro-2700x](https://huggingface.co/datasets/Crownelius/Creative-Writing-Gemini3Pro-2700x)
- [Crownelius/Creative-Writing-Sonnet4.6-800x](https://huggingface.co/datasets/Crownelius/Creative-Writing-Sonnet4.6-800x)
- [OpenMed/Medical-Reasoning-SFT-Mega](https://huggingface.co/datasets/OpenMed/Medical-Reasoning-SFT-Mega)
- [Jackrong/glm-4.7-multiturn-CoT](https://huggingface.co/datasets/Jackrong/glm-4.7-multiturn-CoT)
- [ianncity/KIMI-K2.5-1000000x](https://huggingface.co/datasets/ianncity/KIMI-K2.5-1000000x)
- [ianncity/Hunter-Alpha-SFT-300000x](https://huggingface.co/datasets/ianncity/Hunter-Alpha-SFT-300000x)
- [ianncity/Hunter-Alpha-Programming-160000x](https://huggingface.co/datasets/ianncity/Hunter-Alpha-Programming-160000x)
- [vanty120/Gpt-5.4-Xhigh-Reasoning-2000x](https://huggingface.co/datasets/vanty120/Gpt-5.4-Xhigh-Reasoning-2000x)
- [angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k](https://huggingface.co/datasets/angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k)
- [WithinUsAI/GPT5.5_thinking_max_distill_god_seed_25K](https://huggingface.co/datasets/WithinUsAI/GPT5.5_thinking_max_distill_god_seed_25K)
- [Roman1111111/gpt5.5-terminal](https://huggingface.co/datasets/Roman1111111/gpt5.5-terminal)
- [Jackrong/Claude-opus-4.7-TraceInversion-5000x](https://huggingface.co/datasets/Jackrong/Claude-opus-4.7-TraceInversion-5000x)
- [Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned](https://huggingface.co/datasets/Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned)
- [Jackrong/DeepSeek-V4-Distill-8000x](https://huggingface.co/datasets/Jackrong/DeepSeek-V4-Distill-8000x)
- [guell00/Astra-COT-40K](https://huggingface.co/datasets/guell00/Astra-COT-40K)