datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
arc-agi3-kimi-k2.7-ar25
ARC-AGI-3 ar25 — Agent Trajectories (kimi-k2.7)
Gameplay trajectories from the harness×model pair kimi-k2.7 playing the
ARC-AGI-3 game ar25, part of the
ARA-as-world-model generalization experiment. The agent builds a structured world model
(an Agent-Native Research Artifact) live during play and consults it to crack levels it
cannot solve from cold exploration.
One dataset repo per harness×model×game: sibling repos
arc-agi3-<harness>-<model>-<game> hold the same game played by… See the full description on the dataset page: https://huggingface.co/datasets/AgentNativeResearchLab/arc-agi3-kimi-k2.7-ar25.kimi-k2.6-claude-code-tracesThis dataset was generated using teich by TeichAI
Prepare these datasets for supervised fine-tuning in just a few lines of code — see the Conversion section below.
Kimi K2.6 Claude Code Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by moonshotai/kimi-k2.6.
JSONL files: 36
Format
Each file is newline-delimited JSON representing a single captured agent session.
The trace schema is designed for… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/kimi-k2.6-claude-code-traces.Kimi-K2.5-Reasoning-1M-Cleaned
🪐 Kimi-K2.5-Reasoning-1M-Cleaned
Kimi-K2.5-Reasoning-1M-Cleaned is a cleaned derivative of ianncity/KIMI-K2.5-1000000x. It preserves the original four-config layout from the source dataset and rewrites each record into a unified reasoning-SFT schema with id, conversations, input, output, domain, and meta.
Summary
Source dataset: ianncity/KIMI-K2.5-1000000x
Source author: ianncity
Teacher model recorded in meta.teacher_model: KIMI-K2.5
Token lengths computed with… See the full description on the dataset page: https://huggingface.co/datasets/Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned.DeepSWE-Agent-Kimi-K2-Trajectories-2.8KKIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/ianncity/KIMI-K2.5-1000000x.Creative-Writing-KimiK2.5-Cleaned
Creative-Writing-KimiK2.5-Cleaned
Cleaned creative writing SFT dataset from Kimi K2.5 (655 samples). Prompts cleaned, thinking traces preserved.
Format
Each line is a JSON object with:
messages: list of message dicts with roles (system, user, assistant)
System: writing quality instructions
User: cleaned creative writing prompt
Assistant: creative writing response (may include <think> traces)
Stats
Metric
Value
Total prompt tokens
80… See the full description on the dataset page: https://huggingface.co/datasets/Crownelius/Creative-Writing-KimiK2.5-Cleaned.DeepSWE-Agent-Kimi-K2-Trajectories-Rejection-SamplingKimi-K2.5-Reasoning-General-Sharded
Kimi-K2.5-Reasoning-General-Sharded
Byte-preserving sequential 100 MB JSONL shards of selected files from Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned. All credit for data generation and upstream curation belongs to the source authors. See the upstream dataset card for attribution, source descriptions and license terms.
Included files: General-Distillation.jsonl.
No filtering, shuffling, normalization, tokenization or truncation was performed. Complete records and all original fields… See the full description on the dataset page: https://huggingface.co/datasets/placeholderlabs/Kimi-K2.5-Reasoning-General-Sharded.kimi-k2.6-agentThis dataset was generated using teich by TeichAI
Prepare these datasets for supervised fine-tuning in just a few lines of code — see the Conversion section below.
Kimi K2.6 Agent Traces
This directory contains raw agent trace files generated by teich.
All assistant responses were generated by moonshotai/kimi-k2.6.
JSONL files: 15
Training-ready tools
A complete configured tools schema snapshot is embedded in the collapsed section at the bottom of this README.… See the full description on the dataset page: https://huggingface.co/datasets/armand0e/kimi-k2.6-agent.Kimi-K2.5-Reasoning-1M-Cleaned
🪐 Kimi-K2.5-Reasoning-1M-Cleaned
Kimi-K2.5-Reasoning-1M-Cleaned is a cleaned derivative of ianncity/KIMI-K2.5-1000000x. It preserves the original four-config layout from the source dataset and rewrites each record into a unified reasoning-SFT schema with id, conversations, input, output, domain, and meta.
Summary
Source dataset: ianncity/KIMI-K2.5-1000000x
Source author: ianncity
Teacher model recorded in meta.teacher_model: KIMI-K2.5
Token lengths computed with… See the full description on the dataset page: https://huggingface.co/datasets/rAVEUK/Kimi-K2.5-Reasoning-1M-Cleaned.Creative-Writing-Reasoning-KimiK2.5-600x
Pulitzer Diamond Prose KIMI Seeds
This dataset contains 655 high-quality creative writing seeds generated using Kimi-v1.
Each entry represents a story opening designed to meet high literary standards, including internal thinking traces used during generation.
How it was made
The data was generated using a custom multi-platform generation engine. Models were prompted with a specialized "Diamond Quality" seed template that enforces strict literary requirements:… See the full description on the dataset page: https://huggingface.co/datasets/Crownelius/Creative-Writing-Reasoning-KimiK2.5-600x.Kimi-K2.7-CodingTraces-9000x
Kimi K2.7 Coding Traces 9000x
A validated 9,014-row coding and software-engineering reasoning
dataset generated with moonshotai/Kimi-K2.7-Code.
Every row contains a coding-focused prompt, a separated reasoning trace, and a
final answer. The release was built from a durable Google Drive generation
pipeline and underwent a complete two-pass schema and delimiter audit before
publication.
Generation configuration
Setting
Value
Teacher… See the full description on the dataset page: https://huggingface.co/datasets/trjxter/Kimi-K2.7-CodingTraces-9000x.Kimi-K2.5-Reasoning-1M-Cleaned
🪐 Kimi-K2.5-Reasoning-1M-Cleaned
Kimi-K2.5-Reasoning-1M-Cleaned is a cleaned derivative of ianncity/KIMI-K2.5-1000000x. It preserves the original four-config layout from the source dataset and rewrites each record into a unified reasoning-SFT schema with id, conversations, input, output, domain, and meta.
Summary
Source dataset: ianncity/KIMI-K2.5-1000000x
Source author: ianncity
Teacher model recorded in meta.teacher_model: KIMI-K2.5
Token lengths computed… See the full description on the dataset page: https://huggingface.co/datasets/JBrightmanAI/Kimi-K2.5-Reasoning-1M-Cleaned.kimi-k25-tiny-fidelity-root-v1
kimi-k25 random CPU fixture root
A root fidelity dataset in hidden form, produced by engines/tools/hf_capture.py from malaiwah/kimi-k25-tiny-random-bf16.
The cut
the final hidden state handed to lm_head -- after the text model's final norm and immediately before the head matmul -- captured as the head module's input via torch.nn.Module.register_forward_pre_hook; replay applies the head ONLY (no final norm at replay time: the capture already sits after it). Same… See the full description on the dataset page: https://huggingface.co/datasets/malaiwah/kimi-k25-tiny-fidelity-root-v1.Code-sonnet-5-gpt-5.5-kimi-k2.5Hello guys, this is a dataset from an AI that's good at writing code uh...
Kimi-K2.6-Reasoning-3300x-WandB
Kimi-K2.6-Reasoning-3300x-WandB
Kimi-K2.6-Reasoning-3300x-WandB is a W&B-only synthetic reasoning dataset generated with Kimi-K2.6 through Weights & Biases Inference.
This dataset is the pure W&B-generated subset from a larger planned 8,000-example Kimi reasoning distillation run. Generation stopped when the W&B quota limit was reached, and the completed accepted rows were audited, cleaned, and exported as a standalone dataset.
This release contains 3,303 accepted W&B-generated rows… See the full description on the dataset page: https://huggingface.co/datasets/trjxter/Kimi-K2.6-Reasoning-3300x-WandB.KIMI-K2.5-1000000-RU
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/YurinKO/KIMI-K2.5-1000000-RU.KimiK2.5-2000x
Kimi K2.5 9000x Dataset
Dataset Description
This dataset contains 2144 high-quality samples generated using Kimi K2.5 model, covering diverse tasks including code generation, mathematical reasoning, and general problem-solving.
Dataset Summary
Total Samples: 2144
Model: Kimi K2.5
Languages: English
Format: JSON
License: Apache 2.0
Task Distribution
The dataset includes samples across multiple domains:
Code Generation: Programming… See the full description on the dataset page: https://huggingface.co/datasets/Crownelius/KimiK2.5-2000x.KIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/nick007x/KIMI-K2.5-1000000x.Kimi-K2.5-Reasoning-Reduced-Luna
Kimi K2.5 Reasoning Reduced with GPT-5.6 Luna
This dataset contains synthetic, lossy compressions of reasoning traces from
Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned,
configuration General-Distillation.
The final-answer suffix is copied programmatically from the source and is not
regenerated by the model. The compressed reasoning is synthetic and is not
guaranteed to preserve every logical detail.
Training columns
Train on conversations_reduced or output_reduced. The… See the full description on the dataset page: https://huggingface.co/datasets/Miska25/Kimi-K2.5-Reasoning-Reduced-Luna.kimi-k2-thinking-1000xThis is a reasoning dataset created using Kimi k2 thinking from MoonshotAI. Some of these questions are from reedmayhew and the rest were generated.
Most of the questions cover the following topics: Web Development, Logic, Math, Embedded Systems, Web Design and Python Scripting.
The dataset is meant for creating distilled versions of Kimi k2 thinking by fine-tuning already existing open-source LLMs.
kimi-k2.5-reasoning-1m-cleaned
🪐 Kimi-K2.5-Reasoning-1M-Cleaned
Kimi-K2.5-Reasoning-1M-Cleaned is a cleaned derivative of ianncity/KIMI-K2.5-1000000x. It preserves the original four-config layout from the source dataset and rewrites each record into a unified reasoning-SFT schema with id, conversations, input, output, domain, and meta.
Summary
Source dataset: ianncity/KIMI-K2.5-1000000x
Source author: ianncity
Teacher model recorded in meta.teacher_model: KIMI-K2.5
Token lengths computed… See the full description on the dataset page: https://huggingface.co/datasets/ansulev/kimi-k2.5-reasoning-1m-cleaned.Kimi-K2.6-Thinking-200x
Dataset Card (Kimi-K2.6-Thinking-200x)
Dataset Summary
Kimi-K2.6-Reasoning-207 is a high-quality distilled reasoning dataset designed for supervised fine-tuning (SFT) of small language models.
This dataset uses a curated seed question set covering Mathematics, Code, Logic, Science, Analysis, and Instruction-following domains. By calling the Kimi-K2.6 model via the Moonshot AI API as the teacher model, it generates high-quality responses featuring long-form step-by-step… See the full description on the dataset page: https://huggingface.co/datasets/uniquealexx/Kimi-K2.6-Thinking-200x.Kimi-K2.6-Technical-Reasoning-AddOn-3300x
Kimi-K2.6-Technical-Reasoning-AddOn-3300x
This dataset is a technical reasoning add-on dataset generated with Kimi K2.6 as the teacher model.
The dataset was designed as an additional technical reasoning trace set for downstream SFT experiments, especially around math, graduate-level science, coding, and debugging/code-repair style prompts.
Dataset Summary
Dataset name: Kimi-K2.6-Technical-Reasoning-AddOn-3300x
Teacher model: Kimi-K2.6
Backend: W&B… See the full description on the dataset page: https://huggingface.co/datasets/trjxter/Kimi-K2.6-Technical-Reasoning-AddOn-3300x.Kimi-K2.5-Reasoning-1M-Cleaned
🪐 Kimi-K2.5-Reasoning-1M-Cleaned
Kimi-K2.5-Reasoning-1M-Cleaned is a cleaned derivative of ianncity/KIMI-K2.5-1000000x. It preserves the original four-config layout from the source dataset and rewrites each record into a unified reasoning-SFT schema with id, conversations, input, output, domain, and meta.
Summary
Source dataset: ianncity/KIMI-K2.5-1000000x
Source author: ianncity
Teacher model recorded in meta.teacher_model: KIMI-K2.5
Token lengths computed with… See the full description on the dataset page: https://huggingface.co/datasets/EngMuhammadAtef/Kimi-K2.5-Reasoning-1M-Cleaned.KIMI-K2.5-filteredKIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/WWX0825/KIMI-K2.5-1000000x.kimi-k2-0905-20M
20M token synthetic instruction dataset (Kimi 0905)
User prompts are extracted from three curated instruction-following datasets. Low-quality and repetitive prompts are identified and removed or rewritten using Gemini 3 Flash (+adding medatada for each message). The resulting 15,825 filtered user prompts are sent to Kimi K2 0905 to generate high-quality synthetic responses.
Difficulty Split
Medium: 48.3% (7,638)
Hard: 27.6% (4,373) — mostly from… See the full description on the dataset page: https://huggingface.co/datasets/xrist0bg/kimi-k2-0905-20M.KIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/bitsydarel/KIMI-K2.5-1000000x.KIMI-K2.5-1000000x
KIMI-K2.5-1000000x
1,000,000 reasoning traces distilled from KIMI-K2.5 on high reasoning, (Each subset has different questions)
Distribution:
Coding: 50% (Includes: Webdev, Python, C++, Java, JS, C, Ruby, Lua, Rust, and C#)
Science: 20% (Physics, Chemistry, Biology) - 100k more completions in the PHD-Science subset
Math: 15% (Algebra, Calculus, Probability) - 200k more completions in kimiMath200k.jsonl
Computer Science: 5%
Logical Questions: 5%
Creative Writing: 5%… See the full description on the dataset page: https://huggingface.co/datasets/TheDrMoniker/KIMI-K2.5-1000000x.