Model Architecture: Gemma 4
- Input: Text / Image / Audio
- Output: Text
Model Optimizations:
- Weight quantization: FP4
- Activation quantization: None (16-bit)
Release Date: 2026-07-03
Version: 1.0
Quantized by: xdavxd
Base Model: coder3101/gemma-4-E2B-it-heretic
Original Model: google/gemma-4-E2B-it

This model is a quantized version of coder3101/gemma-4-E2B-it-heretic. It was evaluated on several tasks to assess its quality in comparison to the original model.

Model Optimizations

This model was obtained by quantizing the weights and activations of coder3101/gemma-4-E2B-it-heretic to FP4 data type using the NVFP4 format, ready for inference with vLLM. This optimization reduces the number of bits per parameter from 16 to 4, reducing the disk size and GPU memory requirements by approximately 25%.

Weights are quantized with FP4 (group_size=16) using local per-group scaling. Only the weights of the linear operators within transformer blocks are quantized using LLM Compressor. Vision, audio, embedding, and output head layers are kept in their original precision.

Deployment

Use with vLLM

This model can be deployed using vLLM. For detailed instructions including multi-GPU deployment, multimodal inference, thinking mode, function calling, and benchmarking, see the Gemma 4 vLLM usage guide.

Start the vLLM server:

vllm serve xdavxd/gemma-4-E2B-it-heretic-NVFP4A16 \
  --max-model-len 32768 \
  --gpu-memory-utilization 0.50

To enable thinking/reasoning and tool calling:

vllm serve xdavxd/gemma-4-E2B-it-heretic-NVFP4A16 \
  --max-model-len 32768 \
  --gpu-memory-utilization 0.50 \
  --enable-auto-tool-choice \
  --reasoning-parser gemma4 \
  --tool-call-parser gemma4 \
  --chat-template examples/tool_chat_template_gemma4.jinja \
  --limit-mm-per-prompt '{"image": 4, "audio": 1}' \
  --async-scheduling

Tip: For text-only workloads, pass --limit-mm-per-prompt '{"image": 0, "audio": 0}' to skip vision encoder memory allocation and free up GPU memory for a longer context window.

Send requests to the server:

python

from openai import OpenAI

openai_api_key = "EMPTY"
openai_api_base = "http://<your-server-host>:8001/v1"

client = OpenAI(
    api_key=openai_api_key,
    base_url=openai_api_base,
)

model = "xdavxd/gemma-4-E2B-it-heretic-NVFP4A16"

messages = [
    {"role": "user", "content": "Explain quantum mechanics clearly and concisely."},
]

outputs = client.chat.completions.create(
    model=model,
    messages=messages,
)

generated_text = outputs.choices[0].message.content
print(generated_text)

Creation

This model was created by applying NVFP4 quantization with LLM Compressor, as presented in the code snippet below.

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from llmcompressor import oneshot
from llmcompressor.modifiers.quantization import QuantizationModifier

MODEL_ID = "./gemma-4-E2B-it-heretic"
SAVE_DIR = "gemma-4-E2B-it-heretic-NVFP4A16"

model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

recipe = QuantizationModifier(
    targets="Linear",
    scheme="NVFP4A16",
    ignore=["lm_head", "re:.*embed.*", "re:.*vision_tower.*", "re:.*audio_tower.*", "re:.*per_layer.*"],
)

oneshot(
    model=model,
    tokenizer=tokenizer,
    recipe=recipe,
)

model.save_pretrained(SAVE_DIR, save_compressed=True)
tokenizer.save_pretrained(SAVE_DIR)

Evaluation

This model was evaluated on GSM8K Platinum, IFEval, and MATH-500, using lm-evaluation-harness and lighteval, served with vLLM (OpenAI-compatible API). All evaluations were performed with thinking enabled.

Performance

(as reported by coder3101)

Metric	This model	Original model (google/gemma-4-E2B-it)
KL divergence	0.1651	0 (by definition)
Refusals	5/100	98/100

Also, see the Gemma 4 E2B: 13 Abliteration Techniques Compared report by DreamFast. coder3101 was rated as "Best overall" and noted it as "best capability-to-safety tradeoff in the comparison."

Accuracy

Category	Benchmark	google/gemma-4-E2B-it	xdavxd/gemma-4-E2B-it-heretic-NVFP4A16	Recovery
Instruction Following	IFEval (0-shot, prompt-level strict)	-	-	-
IFEval (0-shot, inst-level strict)	-	-	-
Reasoning	GSM8K Platinum (flexible-extract)	-	-	-
GSM8K Platinum (strict-match)	-	-	-
MATH-500 (0-shot, pass@1)	-	-	-

Reproduction

The results were obtained using the following commands:

Each benchmark was run 3 times with different judgmental seeds (1234, 2345, 3456) and the scores were averaged; AIME 2025 used 8 seeds.

Ran using a docker image, you will need to tweak these for your own environment. Kept parameters similar to the RedHatAI runs.

vLLM server (all benchmarks):

sudo docker run --rm -it \
  --gpus all \
  --ipc=host \
  --network host \
  -v ~/quant-workspace/gemma-4-E2B-it-heretic-NVFP4A16:/models/heretic:ro \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  -e VLLM_USE_V2_MODEL_RUNNER=1 \
  ghcr.io/timothystewart6/vllm-gb10:latest \
  vllm serve /models/heretic \
  --host 0.0.0.0 --port 8001 \
  --served-model-name heretic \
  --max-model-len 69632 \
  --gpu-memory-utilization 0.50 \
  --enable-auto-tool-choice \
  --reasoning-parser gemma4 \
  --tool-call-parser gemma4 \
  --limit-mm-per-prompt '{"image":0,"audio":0}' \
  --kv-cache-dtype fp8 \
  --max-num-seqs 32 \
  --async-scheduling

GSM8K Platinum (lm-eval, 0-shot, 3 repetitions)

for SEED in 1234 2345 3456; do
  lm_eval --model local-chat-completions \
    --tasks gsm8k_platinum_cot_llama \
    --model_args "model=heretic,max_length=36096,base_url=http://0.0.0.0:8001/v1/chat/completions,num_concurrent=32,max_retries=3,tokenized_requests=False,tokenizer_backend=None,timeout=2400" \
    --num_fewshot 0 \
    --apply_chat_template \
    --output_path "results_gsm8k_seed${SEED}.json" \
    --seed $SEED \
    --gen_kwargs "do_sample=True,temperature=1.0,top_p=0.95,top_k=64,max_gen_toks=32000,seed=${SEED}"
  echo "Seed $SEED complete"
done

IFEval (lm-eval, 0-shot, 3 repetitions)

for SEED in 1234 2345 3456; do
  lm_eval --model local-chat-completions \
    --tasks ifeval \
    --model_args "model=heretic,max_length=36096,base_url=http://0.0.0.0:8001/v1/chat/completions,num_concurrent=32,max_retries=3,tokenized_requests=False,tokenizer_backend=None,timeout=2400" \
    --num_fewshot 0 \
    --apply_chat_template \
    --output_path "results_ifeval_seed${SEED}.json" \
    --seed $SEED \
    --gen_kwargs "do_sample=True,temperature=1.0,top_p=0.95,top_k=64,max_gen_toks=32000,seed=${SEED}"
  echo "IFEval seed $SEED complete"
done

MATH-500, (lighteval, 3 repetitions)

litellm_config.yaml:

yaml

model_parameters:
  provider: hosted_vllm
  model_name: hosted_vllm/heretic
  base_url: http://0.0.0.0:8001/v1
  api_key: ''
  timeout: 3600
  concurrent_requests: 32
  generation_parameters:
    temperature: 1.0
    max_new_tokens: 65536
    top_p: 0.95
    top_k: 64
    seed: 1234

# MATH-500 (3 seeds)
for SEED in 1234 2345 3456; do
  sed -i "s/seed: .*/seed: $SEED/" litellm_config.yaml
  lighteval endpoint litellm litellm_config.yaml 'math_500|0' \
    --output-dir "results_math500_seed${SEED}/" --save-details
  echo "MATH-500 seed $SEED complete"
done