Skip to content
Language models

WeightSeek-Format22-0.8B

A specialized LLM for JSON tasks

2 min readupdated

WeightSeek-Format22-0.8B is a standalone 0.75-billion-parameter text model (LLM, language model) in FP16 format (half-precision computing), built for experimenting with precise instruction following. Its key advantage over the base version is the ability to consistently output a valid JSON object that parsers can read without errors. The model is built upon the work of Rikiya Takehi and colleagues for Qwen3.5-0.8B using weight modifications. It is useful for developers who need a compact local model for structured data output.

Key features

  • Based on the Qwen3.5-0.8B architecture with weight-space programming modifications.
  • Weighs about 1.50 GB in a single safetensors file containing all 752 million parameters.
  • Improved instruction following: the model outputs clean JSON where the previous version failed.
  • Works out of the box with the standard transformers library in environments like Colab or Kaggle on a single GPU.

What you can do

  • Integrate the model into local data processing pipelines to generate JSON configuration files.
  • Use it as a compact parser in projects where strict response syntax is critical.
  • Conduct experiments with weight-space programming and precise text prompt adherence.

Example

import json
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "efe-T/WeightSeek-Format22-0.8B"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, dtype=torch.float16, device_map="cuda", attn_implementation="eager"
).eval()

messages = [{
    "role": "user",
    "content": "One JSON object is needed for a config file. Set the property 'elmwood' to the string 'almondwood'.",
}]
prompt = tokenizer.apply_chat_template(
    messages, tokenize=False, add_generation_prompt=True, enable_thinking=False
)
inputs = tokenizer(prompt, add_special_tokens=False, return_tensors="pt").to(model.device)
with torch.inference_mode():
    output = model.generate(
        **inputs, do_sample=False, max_new_tokens=None,
        max_length=model.config.max_position_embeddings,
        eos_token_id=[248044, 248046], pad_token_id=tokenizer.pad_token_id,
    )
answer = tokenizer.decode(output[0, inputs.input_ids.shape[1]:], skip_special_tokens=True)
print(json.loads(answer))  # {'elmwood': 'almondwood'}
Read moreDetails
Benchmarks: on a test of 105 prompts, the model scored 50 points compared to 42 for the parent version (including 8 successful responses in Plain JSON format instead of 0).
Hardware requirements: runs on a standard GPU (such as Colab T4) thanks to FP16 and a size of 1.50 GB.
Limitations: JSON improvements are mostly observed in English (results remained at zero for Turkish prompts). Distributed under the Apache 2.0 license.
Discuss Telegram 0 15

Will it run?

Runs in F16

Fits without quantisation.

  • F16weights 1.4 GB · ready files2.4 GB
  • FP8 / INT8recommendedweights 0.7 GB1.7 GB
  • 4-bitweights 0.4 GB1.4 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · F16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Language models: with a 8192-token context.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

A ready script for the machine we advise above: paste it and the model downloads and starts.

  1. 1. Rent A30

    $0.26 an hour with the coupon. Pick the Ubuntu image when you create the machine.

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # efe-T/WeightSeek-Format22-0.8B — A30, FP8 / INT8
    # NVIDIA driver: CUDA 13 needs 580+. If older or missing, the current nvidia-open is installed (keep yours: KEEP_DRIVER=1)
    DRV=$(nvidia-smi --query-gpu=driver_version --format=csv,noheader 2>/dev/null | head -1 | cut -d. -f1)
    case "$DRV" in ""|*[!0-9]*) DRV=0 ;; esac
    if [ "$DRV" -lt 580 ] && [ -z "${KEEP_DRIVER:-}" ]; then
      . /etc/os-release && \
      curl -fsSLO "https://developer.download.nvidia.com/compute/cuda/repos/ubuntu${VERSION_ID//./}/x86_64/cuda-keyring_1.1-1_all.deb" && \
      sudo dpkg -i cuda-keyring_1.1-1_all.deb && rm -f cuda-keyring_1.1-1_all.deb && \
      sudo apt-get update && \
      sudo DEBIAN_FRONTEND=noninteractive apt-get install -y nvidia-open || \
      { echo "Could not install the NVIDIA driver: see the error above. To keep the current one, run with KEEP_DRIVER=1."; exit 1; }
      echo "NVIDIA driver updated. Reboot the machine (sudo reboot) and run this same script again."
      exit 0
    fi
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv venv ~/vllm --python 3.12 && source ~/vllm/bin/activate
    uv pip install vllm
    vllm serve efe-T/WeightSeek-Format22-0.8B --max-model-len 8192 --quantization fp8
  3. 3. Use it

    The server listens on port 8000 and speaks the OpenAI API: http://localhost:8000/v1 on the machine. From your computer: ssh -L 8000:localhost:8000 user@machine-IP.

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

  • MTP (multi-token prediction)0.34.0from November 16, 2026new release

    1.5–2× generation speed

    If the model has MTP layers (DeepSeek-V3/R1, Qwen3-Next, GLM, MiMo and others), it predicts several tokens per step itself.

    How to enable: vLLM / SGLang: --speculative-config '{"method": "mtp", "num_speculative_tokens": 1}'. Check the model card.

    Earlier generations (2)
    • EAGLE-3 — Speculative decoding with a small draft head. For models without their own MTP layers.
    • EAGLE-2 / Medusa — Earlier draft-head generations. Still work if that is all you find.
  • FP8 KV cache0.34.0from November 16, 2026new release

    2× more context in the same memory

    Keeps the attention cache in 8 bits, so long contexts fit where they did not.

    How to enable: vLLM: --kv-cache-dtype fp8; llama.cpp: -ctk q8_0 -ctv q8_0.

  • vLLM / SGLang0.34.0from November 16, 2026new release

    many times more tokens/s on a server

    Server engines: batching, FlashAttention, FP8 and speculative decoding out of the box.

    How to enable: vllm serve <model> — an OpenAI-compatible API.

    Earlier generations (1)
    • llama.cpp — GGUF models on any GPU, Mac or even CPU; some layers can stay in RAM.

Model family

  1. Qwen3.5-0.8B-Basefamily root
  2. Qwen3.5-0.8B
  3. Qwen3.5-0.8B-weight-space-programming
  4. WeightSeek-Format22-0.8B

New posts about versions of Qwen3.5-0.8B-Base (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 9All on Hugging Face
Merges 1All on Hugging Face

About the model

efe-T/WeightSeek-Format22-0.8B
Author
efe-T
Category
Language models
Released
Posted

On NeyroLearn

Hugging Face

150

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments