Instructions to use tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic")
model = AutoModelForCausalLM.from_pretrained("tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic

SGLang

How to use tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic with Docker Model Runner:
```
docker model run hf.co/tacodevs/Behemoth-R1-123B-v2-FP8-Dynamic
```

Behemoth-R1-123B-v2-FP8-Dynamic

File size: 1,797 Bytes

de70ea5

---
tags:
- fp8
- quantized
- mistral
- roleplay
- creative-writing
- reasoning
base_model: TheDrummer/Behemoth-R1-123B-v2
library_name: transformers
pipeline_tag: text-generation
license: apache-2.0
---

# Behemoth-R1-123B-v2 FP8 Dynamic

FP8 Dynamic quantization of [TheDrummer/Behemoth-R1-123B-v2](https://huggingface.co/TheDrummer/Behemoth-R1-123B-v2) using llmcompressor.

## Model Details

- **Base Model**: TheDrummer/Behemoth-R1-123B-v2 (Mistral Large 2411 finetune)
- **Quantization**: FP8 Dynamic (W8A8) via llmcompressor
- **Scheme**: FP8_DYNAMIC, lm_head excluded
- **Size**: ~123 GB (vs 246 GB FP16)
- **Format**: SafeTensors with compressed-tensors metadata

## Usage with vLLM

```bash
python3 -m vllm.entrypoints.openai.api_server \
    --model Irvollo/Behemoth-R1-123B-v2-FP8-Dynamic \
    --quantization compressed-tensors \
    --dtype bfloat16 \
    --max-model-len 32768 \
    --gpu-memory-utilization 0.95 \
    --enable-prefix-caching \
    --trust-remote-code
```

## Reasoning / Thinking

Supports native reasoning via `<think>` tag prefill:

```json
{
  "messages": [
    {"role": "user", "content": "Your question"},
    {"role": "assistant", "content": "<think>\n"}
  ],
  "continue_final_message": true,
  "add_generation_prompt": false
}
```

## Hardware Requirements

- **Single GPU**: H200 NVL (141 GB) — tight with ~18 GB KV cache
- **Recommended**: 2x A100 80GB or H100 for comfortable KV headroom

## Quantization Details

- Quantized on 2x NVIDIA B200 (358 GB VRAM)
- Calibration: 616 linear layers in <1 second
- Total pipeline: ~11 minutes
- Tool: [llmcompressor](https://github.com/vllm-project/llm-compressor)

## Credits

- Original model by [TheDrummer](https://huggingface.co/TheDrummer)
- FP8 quantization by [Irvollo](https://huggingface.co/Irvollo)