README.md

---
license: apache-2.0
language:
- en
base_model:
- LLM360/K2-V2
---

# **K2-V2-Instruct**

<img src="https://huggingface.co/LLM360/K2-V2/resolve/main/figures/K2.LOGO.PRIMARY.RGB.png" width="100" alt="K2-V2 model logo"/>

📚 [Tech Report](https://www.llm360.ai/reports/K2_V2_report.pdf) - 📝 [Training Code](https://github.com/llm360/k2v2_train) - 🏢 [Evaluation Code](https://github.com/llm360/eval360) 

🗂️ [Pretraining Data: TxT360](https://huggingface.co/datasets/LLM360/TxT360) - 🗂️ [Midtraining Data: TxT360-Midas](https://huggingface.co/datasets/LLM360/TxT360-Midas) - 🗂️ [SFT Data: TxT360-3efforts](https://huggingface.co/datasets/LLM360/TxT360-3efforts)

K2-V2 is our most capable fully open model to date, and one of the strongest open-weight models in its class. It uses a 70B-parameter dense transformer architecture and represents the latest advancement in the LLM360 model family.


<img src="https://huggingface.co/LLM360/K2-V2/resolve/main/figures/sft-models.png" width="400" alt="K2-V2 SFT results"/>

Beyond standard competencies such as factual knowledge and conversational ability, K2-V2 demonstrates strong long-context consistency, deep mathematical understanding, and robust reasoning skills. These capabilities serve as building blocks for sophisticated downstream applications, such as solving complex math problems and executing agentic workflows.


<img src="https://huggingface.co/LLM360/K2-V2/resolve/main/figures/base-models.png" width="400" alt="K2-V2 GPQA results"/>

---

## **Quick Start**

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("llm360/k2-v2-instruct", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("llm360/k2-v2-instruct")

prompt = "Explain why the derivative of sin(x) is cos(x)."
messages = [
    {"role": "system", "content": "You are K2, a helpful assistant created by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) Institute of Foundation Models (IFM)."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
    reasoning_effort="high" # Or "medium"/"low"
)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=200)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

Alternatively, you may serve the model using VLLM:

```
vllm serve LLM360/K2-V2-Instruct --tensor-parallel-size 8 --port 8000
```

K2-V2-Instruct uses `reasoning_effort="low"|"medium"|"high"` in the chat template to determine reasoning effort. If you cannot use `tokenizer.apply_chat_template`, you may also pass in these arguments using `extra_body` and `chat_template_kwargs`:

```
from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="key"
)

completion = client.chat.completions.create(
    model="LLM360/K2-V2-Instruct",
    messages = [
        {"role": "system", "content": "You are K2, a helpful assistant created by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) Institute of Foundation Models (IFM)."},
        {"role": "user", "content": "Explain why the derivative of sin(x) is cos(x)."}
    ],
    extra_body={
        "chat_template_kwargs": {"reasoning_effort": "high"},
    },
)
```

## With Tool-use
To enable tool-use, load, tokenize, and generate as the examples above, but additionally define a set of tools and pass them to the chat template.

**Using transformers**
```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("llm360/k2-v2-instruct", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("llm360/k2-v2-instruct")

prompt = "Explain why the derivative of sin(x) is cos(x)."
messages = [
    {"role": "system", "content": "You are K2, a helpful assistant created by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) Institute of Foundation Models (IFM)."},
    {"role": "user", "content": prompt}
]

# Define list of tools
tools = [
    {
        "type": "function",
        "name": "get_horoscope",
        "description": "Get today's horoscope for an astrological sign.",
        "parameters": {
            "type": "object",
            "properties": {
                "sign": {
                    "type": "string",
                    "description": "An astrological sign like Taurus or Aquarius",
                },
            },
            "required": ["sign"],
        },
    },
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
    reasoning_effort="high" # Or "medium"/"low"
    tools=tools
)

inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=200)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

**Using vLLM**
Serve the model with tool-use enabled and parser set.

```
vllm serve LLM360/K2-V2-Instruct --enable-auto-tool-choice --tool-call-parser hermes --tensor-parallel-size 8 --port 8000 
```

```
from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8000/v1",
    api_key="key"
)

# Define list of tools
tools = [
    {
        "type": "function",
        "name": "get_horoscope",
        "description": "Get today's horoscope for an astrological sign.",
        "parameters": {
            "type": "object",
            "properties": {
                "sign": {
                    "type": "string",
                    "description": "An astrological sign like Taurus or Aquarius",
                },
            },
            "required": ["sign"],
        },
    },
]

completion = client.chat.completions.create(
    model="LLM360/K2-V2-Instruct",
    messages = [
        {"role": "system", "content": "You are K2, a helpful assistant created by Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) Institute of Foundation Models (IFM)."},
        {"role": "user", "content": "Explain why the derivative of sin(x) is cos(x)."}
    ],
    extra_body={
        "chat_template_kwargs": {"reasoning_effort": "high"},
    },
    tools=tools,
)
```


---

## **Evaluation Summary**

Below we report performance across general, reasoning, mathematical, and coding benchmarks. Scores for K2-V2 checkpoints (base → mid-4) demonstrate the impact of staged mid-training on reasoning quality.

| Task / Model | base | mid-1 | mid-2 | mid-3 | mid-4 | Qwen2.5-72B | Llama3.0-70B | Llama3.1-70B | Olmo3-32B |
|--------------|------|-------|-------|-------|-------|--------------|---------------|---------------|------------|
| **General Tasks** | | | | | | | | | |
| **MMLU** | 74.3 | 74.4 | 73.5 | 75.0 | 75.2 | **86.1** | <u>79.5</u> | 79.3 | 75.2 |
| **MMLU-Pro** | 43.7 | 46.8 | 48.1 | **59.8** | 57.0 | <u>58.1</u> | 52.8 | 53.8 | 49.6 |
| **BBH** | 68.4 | 79.8 | 81.1 | 82.2 | <u>83.2</u> | **86.3** | 82.2 | 82.1 | 77.6 |
| **HELLASWAG** | <u>87.8</u> | 86.9 | 86.6 | 86.6 | 86.0 | 87.6 | **88.0** | 85.0 | 84.8 |
| **WINOGRANDE** | 82.6 | 83.7 | 83.7 | 83.7 | 83.0 | 83.9 | <u>85.3</u> | 79.8 | **90.3** |
| **PIQA** | 84.2 | 84.0 | 83.3 | 82.9 | 83.1 | 83.5 | <u>84.6</u> | 84.3 | **85.6** |
| **TRUTHFULQA** | 54.0 | 54.9 | 55.1 | <u>55.8</u> | 53.9 | **60.5** | 45.6 | 49.7 | 54.9 |
| **Math & STEM Tasks** | | | | | | | | | |
| **GPQA-DIAMOND** | 26.3 | 31.3 | 27.8 | <u>43.9</u> | **55.1** | 34.9 | 21.2 | 27.3 | 30.3 |
| **GSM8K** | 68.0 | 76.4 | 82.1 | **93.6** | <u>92.5</u> | 91.2 | 83.2 | 81.1 | 80.5 |
| **MATH** | 27.8 | 38.2 | 41.1 | **94.7** | <u>91.4</u> | 58.5 | 41.9 | 41.6 | 43.4 |
| **AIME 2025** | 0.0 | 17.6 | 25.1 | **53.2** | <u>46.9</u> | 1.7 | 0.1 | 0.2 | 14.7 |
| **ARC-CHALLENGE** | 64.9 | 66.4 | 66.4 | 66.0 | 66.3 | **72.4** | <u>69.2</u> | 64.9 | 65.4 |
| **Coding Tasks** | | | | | | | | | |
| **MBPP** | 57.6 | 57.8 | 58.2 | 59.8 | 61.8 | **75.4** | <u>69.2</u> | 64.4 | 60.2 |
| **HUMANEVAL** | 50.0 | 51.2 | <u>53.7</u> | **54.3** | **54.3** | **54.3** | 42.1 | 50.6 | 36.0 |
| **Logic Puzzles** | | | | | | | | | |
| **COUNTDOWN** | 1.3 | <u>53.3</u> | 53.1 | 35.9 | **75.6** | 6.0 | 1.0 | 0.5 | 23.2 |
| **KK-4 PEOPLE** | 4.8 | 44.9 | <u>68.0</u> | 64.5 | **92.9** | 26.1 | 4.2 | 7.6 | 42.4 |
| **KK-8 PEOPLE** | 0.5 | 23.2 | 41.3 | <u>51.6</u> | **82.8** | 5.7 | 1.1 | 1.3 | 13.0 |
| **ORDER-15 ITEMS** | 4.7 | 30.7 | 47.2 | <u>55.8</u> | **87.6** | 37.0 | 3.5 | 4.5 | 25.0 |
| **ORDER-30 ITEMS** | 0.0 | 0.3 | 3.0 | <u>34.1</u> | **40.3** | 0.7 | 0.2 | 0.1 | 0.6 |
| **Instruction Following** | | | | | | | | | |
| **IFEVAL** | 17.4 | 26.2 | 28.5 | <u>34.5</u> | 26.7 | **40.3** | 15.1 | 17.4 | 13.2 |
| **Arabic** | | | | | | | | | |
| **MMLU-Arabic** | 65.4 | 66.1 | 64.5 | 66.6 | 65.5 | **74.1** | 65.0 | <u>66.8</u> | 47.8 |


Below we report the evaluation results for K2-V2 after supervised fine-tuning (SFT). These variants correspond to three levels of reasoning effort (Low < Medium < High).

| Metric / Model | **K2 Low**<br><sub>Dense · 70B</sub> | **K2 Medium**<br><sub>Dense · 70B</sub> | **K2 High**<br><sub>Dense · 70B</sub> | **Olmo3 Think SFT**<br><sub>Dense · 32B · No RL</sub> | **Olmo3 Think**<br><sub>Dense · 32B · RL</sub> | **GLM-4.5 Air**<br><sub>MoE · 106B A12B</sub> | **MiniMax-M2**<br><sub>MoE · 230B A10B</sub> | **Qwen3 235B**<br><sub>MoE · 235B A22B · Reasoning</sub> | **Qwen 2.5 72B**<br><sub>Dense · 72B</sub> |
|--------|--------------------------------------|------------------------------------------|----------------------------------------|------------------------------------------------------|--------------------------------------------------|----------------------------------------------------|------------------------------------------------------|--------------------------------------------------------------------|-------------------------------------------|
| **LongBench V2** | 40.7 | 41.3 | 42.6 | 42.8 | 47.1 | 49.4 | 55.8 | 60.9 | 47.2 |
| **AIME25** | 27.3 | 62.0 | 80.2 | 68.3 | 73.3 | 81.3 | 75.8 | 88.8 | 15.2 |
| **HMMT25** | 19.0 | 45.6 | 71.4 | 43.3 | 50.83 | 73.3 | 63.5 | 84.2 | 9.79 |
| **GSM8K** | 92.4 | 92.0 | 94.8 | 96.1 | 95.7 | 96.1 | 95.4 | 93.5 | 85.8 |
| **Minerva** | 85.0 | 90.6 | 94.5 | 96.9 | 97.3 | 94.9 | 85.3 | 98.0 | 82.1 |
| **GPQA-D** | 48.5 | 60.6 | 69.3 | 58.0 | 59.8 | 75.3 | 76.2 | 80.7 | 50.5 |
| **MBPP** | 71.0 | 75.8 | 84.8 | 87.6 | 91.6 | 82.8 | 83.8 | 96.2 | 80.0 |
| **HumanEval** | 82.3 | 91.5 | 91.5 | 96.3 | 96.3 | 97.6 | 89.6 | 94.5 | 85.4 |
| **LCBv6** | 39.9 | 51.3 | 67.0 | 67.9 | 67.6 | 67.8 | 79.2 | 72.8 | 36.7 |

Please refer to our [Tech Report](https://www.llm360.ai/reports/K2_V2_report.pdf) for detailed evaluation results.

---

## **Datasets & Mixtures**

### **SFT Mix**

* **TxT360-3efforts**: curated instruction + mixed-difficulty reasoning traces
* Tool-calling demonstrations
* Small but high-value corpus to showcase model potential

All mixtures, filtering rules, and data sources are fully released for reproducibility.

Please refer to our [Tech Report](https://www.llm360.ai/reports/K2_V2_report.pdf) for detailed datasets and mixtures information.

---

## **Model Description**
- **Model type:** K2-V2 follows a standard decoder-only transformer with grouped-query attention and RMSNorm.
- **Training stage:** Pre-training & Post-training
- **Language(s) (NLP):** English
- **License:** Apache 2.0



| Model Hyperparameter      | Value |
| ----------- | ----------- |
| Total Parameters      | 70B       |
| Hidden Size   | 8,192        |
| Intermediate Size (FFN)   | 28,672        |
| Number of Attention Heads   | 64        |
| Number of Layers  | 80        |
| RMSNorm ɛ  | 1e-5        |
| Pre-training Seq Length   | 8,192        |
| Post-training Seq Length   | 524,288        |
| Vocab Size | 250,000 |

---

## Citation

If you use K2-V2-Instruct in your research, please cite the following:

```
@misc{llm360_k2v2_2025,
  title         = {K2-V2: A 360-Open, Reasoning-Enhanced LLM},
  author        = {K2 Team},
  year          = {2025},
}
```