selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

docling-project logo

DeskForge-Gemma4-E4B

docling-project·Instruct·released 5 Oct 2026

Apache 2.07.9B128K ctx ≈5.9 GB
Your GPU
ollama run hf.co/mradermacher/DeskForge-Gemma4-E4B-GGUF:Q4_K_M
All 6 runners ↓

DeskForge-Gemma4-E4B is Gemma 4 E4B (instruction-tuned) fine-tuned on 200K grounding examples from DeskForge-1M. Given a desktop screenshot and a natural-language instruction, it answers with the next action as a pyautogui call whose coordinates are fractions of the screen.

Notify me when

How much VRAM does DeskForge-Gemma4-E4B need?

At Q4_K_M DeskForge-Gemma4-E4B needs about 5.9 GB with an 8K context (5.9 GB at Q4, 17 GB at full precision). Pick your GPU to see what fits and how fast it runs.

--vram
any
--gpus
--quant

Needs: ≈6.3 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.

Change quant, context or KV cache to compare
Advanced: context 8K · KV cache 16-bit
Context
KV cache

Smallest GPU for each quant and context (KV cache 16-bit)

Quantisation4K8K16K32K64K128K
Q3_K_M
Q4_K_M
Q5_K_M
Q6_K
Q8_0
FP16

≤16 GB card24–32 GB48–96 GBmore

QuantisationQualityWeightsKV cacheTotal
Q2_KHeavy lossSmallest, noticeably worse qualityHeavy loss4.1 GB file0.7 GB≈5.4 GB
Q3_K_MNoticeable lossSmall, some quality lossNoticeable loss4.5 GB file0.7 GB≈5.9 GB
Q4_K_MrecommendedSmall lossRecommended default for local useSmall loss4.9 GB file0.7 GB≈6.3 GB
Q5_K_MNear-losslessNear-lossless for most tasksNear-lossless5.3 GB file0.7 GB≈6.8 GB
Q6_KNear-losslessPractically losslessNear-lossless5.8 GB file0.7 GB≈7.2 GB
Q8_0LosslessLossless in practice, twice the size of Q4Lossless7.4 GB file0.7 GB≈9.0 GB
FP16 / BF16LosslessFull precision, as releasedLossless14 GB file0.7 GB≈16 GB

Estimates for 8K tokens of context: weights + KV cache + runtime overhead. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.

How to pick a quant

Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.

  • LosslessF16, Q8_0
  • Near-losslessQ6_K, Q5_K_M
  • Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
  • Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
  • Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K

Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B

How much context fits?

Model quant
KV cache

At 8K context, Q4_K_M needs ≈6.3 GB(4.9 GB weights + 0.7 GB KV cache + 0.8 GB overhead). Pick a GPU above to see how much context fits.

051015201K2K4K8K16K32K64K128KGB

Weights + overheadKV cache

Can the RTX 4060 run DeskForge-Gemma4-E4B? →

Speed vs quality per quant

Smaller quants run faster and fit in less memory, but lose quality.

Q2_KQ3_K_MQ4_K_MQ5_K_MQ6_KQ8_0F16

Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.

Real-world speed of DeskForge-Gemma4-E4B

Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →

No medians yet. Be one of the first 3 to report a setup.

All reports

Run DeskForge-Gemma4-E4B locally

--quantQ4_K_Mdefault; pick your GPU to tailor itmradermacher/DeskForge-Gemma4-E4B-GGUF

Q4_K_M straight from Hugging Facedocs
ollama run hf.co/mradermacher/DeskForge-Gemma4-E4B-GGUF:Q4_K_M

Install Ollama

OpenAI-compatible server with a web UIdocs
llama-server -hf mradermacher/DeskForge-Gemma4-E4B-GGUF:Q4_K_M

Downloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.

Install llama.cpp

Download with the lms CLIdocs
lms get mradermacher/DeskForge-Gemma4-E4B-GGUF@Q4_K_M

Then load it from the LM Studio app or with `lms load`.

Install LM Studio

Serve an OpenAI-compatible APIdocs
vllm serve docling-project/DeskForge-Gemma4-E4B
Docker Compose (NVIDIA)docs
yaml
services:
  vllm:
    image: vllm/vllm-openai:latest
    command: docling-project/DeskForge-Gemma4-E4B
    ports:
      - "8000:8000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

API on port 8000.

Install vLLM

Serve an OpenAI-compatible APIdocs
sglang serve docling-project/DeskForge-Gemma4-E4B --host 0.0.0.0 --port 30000
Docker Compose (NVIDIA)docs
yaml
services:
  sglang:
    image: lmsysorg/sglang:latest
    command: sglang serve docling-project/DeskForge-Gemma4-E4B --host 0.0.0.0 --port 30000
    ports:
      - "30000:30000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

API on port 30000.

Install SGLang

Ask about an imagedocs
python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="docling-project/DeskForge-Gemma4-E4B", dtype="auto", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
            {"type": "text", "text": "Describe this image."},
        ],
    }
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])

Install Transformers

  • Jancommunity · desktop app
  • KoboldCpp logoKoboldCppcommunity · desktop app
  • LocalAI logoLocalAIcommunity · inference server

Quantised versions and downloads

About DeskForge-Gemma4-E4B

DeskForge-Gemma4-E4B is Gemma 4 E4B (instruction-tuned) fine-tuned on 200K grounding examples from DeskForge-1M. Given a desktop screenshot and a natural-language instruction, it answers with the next action as a pyautogui call whose coordinates are fractions of the screen. Live demo · Project page · Dataset · Code · Paper · Also available: DeskForge-Qwen3.5-4B Grounding on DeskForge-1M held-out conditions (accuracy, %). New Scenes are unseen scenes with seen attributes; Theme, App and Resolution hold an appearance preset, three applications and a display resolution out of training entirely. Grounding on external GUI benchmarks (accuracy, %), none of which is used for training: Long-horizon task completion (tasks solved). A fixed Qwen3.6-27B planner decides each step and the action model locates its targets, so only the action model differs between the rows:

Full model card on Hugging Face →

Alternatives to DeskForge-Gemma4-E4B

All alternatives →
Qwen logo

Qwen3.5-9B

Qwen·Vision

≈7.2 GB

Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance.

9.7B256K ctx Apache 2.0
6.3MFeb 2026ollama
Qwen logo≈6.8 GB

Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.

8.8B256K ctx Apache 2.0
6MOct 2025ollama
Qwen logo≈5.8 GB

In the past five months since Qwen2-VL’s release, numerous developers have built new models on the Qwen2-VL vision-language models, providing us with valuable feedback. During this period, we focused on building more useful vision-language models.

8.3B125K ctx Apache 2.0
5.3MJan 2025ollama
Zyphra logo

ZAYA1-VL-8B

Zyphra·Vision

≈6.6 GB

ZAYA1-VL-8B is a vision-language model (VLM) built upon Zyphra's ZAYA1-8B LLM. It has state-of-the-art performance among VLMs for its size and inference efficiency.

9.7B32K ctx Apache 2.0
339KApr 2026gguf
black-forest-labs logo

flux-3-image-klein-pup

black-forest-labs·Vision

≈7.1 GB

FLUX 3 Image Klein Prompt Upsampler is a 9B vision-language model that expands a short image request into the detailed prompt FLUX 3 Image Klein works best with. It takes the request, up to 10 reference images and an optional list of bounding boxes, and returns one prompt.

9.4B256K ctx Apache 2.0
38.7KOct 2026gguf
Qwen logo

Qwen3.5-9B-Base

Qwen·Vision

≈7.2 GB

Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance.

9.7B256K ctx Apache 2.0
266.5KFeb 2026gguf

DeskForge-Gemma4-E4B FAQ

How much VRAM does DeskForge-Gemma4-E4B need?
About 5.9 GB at Q4_K_M with an 8K context. Q8_0 needs about 9.4 GB and full precision about 17 GB. Longer contexts need more memory for the KV cache.
Can DeskForge-Gemma4-E4B run on the RTX 4060?
Yes. It fits at Q6_K (about 7.2 GB), at an estimated 23 tokens/s. It fits comfortably on the Apple M4 (16 GB), Radeon RX 6700 XT, RTX 3060 12GB.
Can I use DeskForge-Gemma4-E4B commercially?
Yes. It is released under Apache 2.0. Permissive: use, modify and sell freely; keep the notice and licence text.
What is the context length of DeskForge-Gemma4-E4B?
131,072 tokens (128K). The full window needs far more memory than an 8K context.
How do I run DeskForge-Gemma4-E4B with Ollama or llama.cpp?
With Ollama: ollama run hf.co/mradermacher/DeskForge-Gemma4-E4B-GGUF:Q4_K_M. With llama.cpp: llama-server -hf mradermacher/DeskForge-Gemma4-E4B-GGUF:Q4_K_M, which starts an OpenAI-compatible server.

Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.