Instructions to use EldanRing/Winnow-E4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use EldanRing/Winnow-E4B with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf EldanRing/Winnow-E4B:BF16
# Run inference directly in the terminal:
llama cli -hf EldanRing/Winnow-E4B:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf EldanRing/Winnow-E4B:BF16
# Run inference directly in the terminal:
llama cli -hf EldanRing/Winnow-E4B:BF16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf EldanRing/Winnow-E4B:BF16
# Run inference directly in the terminal:
./llama-cli -hf EldanRing/Winnow-E4B:BF16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf EldanRing/Winnow-E4B:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf EldanRing/Winnow-E4B:BF16

Use Docker

docker model run hf.co/EldanRing/Winnow-E4B:BF16

LM Studio
Jan

vLLM

How to use EldanRing/Winnow-E4B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "EldanRing/Winnow-E4B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "EldanRing/Winnow-E4B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/EldanRing/Winnow-E4B:BF16

Ollama
How to use EldanRing/Winnow-E4B with Ollama:
```
ollama run hf.co/EldanRing/Winnow-E4B:BF16
```
Unsloth Desktop
Docker Model Runner
How to use EldanRing/Winnow-E4B with Docker Model Runner:
```
docker model run hf.co/EldanRing/Winnow-E4B:BF16
```

Lemonade

How to use EldanRing/Winnow-E4B with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull EldanRing/Winnow-E4B:BF16

Run and chat with the model

lemonade run user.Winnow-E4B-BF16

List all available models

lemonade list

Atomic Chat

EldanRing commited on 27 days ago

Commit

734302f

verified ·

1 Parent(s): ba9554e

Update E4B comparison figure title

Browse files

Files changed (2) hide show

README.md +1 -1
assets/e4b-profiles.png +2 -2

README.md CHANGED Viewed

@@ -112,7 +112,7 @@ Download Q8 for the measured decision, chat, and vision setup; add the projector
 Capacity suggestions come from one RTX 5070 Ti using Q8 KV cache, four decision branches, one chat slot, and exclusive memory scheduling. They have not been tested on every GPU in those tiers. The Q8 model and projector answered six questions over a 65,023-position state that included a synthetic image; repeated cached requests took 129–131 ms after an 11.59 s cold request. A separate 62,431-position multimodal chat prompt generated 512 tokens at 71.2 output tokens/s. The [benchmark report](docs/BENCHMARKS.md#64k-context-and-vision-smoke-test) describes these operational probes and their limits.
-![Winnow-E4B and Winnow-12B Q8 GPU memory and chat generation, plus E4B cached 64K decisions](assets/e4b-profiles.png)
 *Q8 operational profiles on an RTX 5070 Ti. The 8K text workload was matched; the 64K vision and chat probes used different images and question counts. The 10 GB+ and 12 GB+ capacity suggestions are inferred from measured memory, not tests on those cards.*

 Capacity suggestions come from one RTX 5070 Ti using Q8 KV cache, four decision branches, one chat slot, and exclusive memory scheduling. They have not been tested on every GPU in those tiers. The Q8 model and projector answered six questions over a 65,023-position state that included a synthetic image; repeated cached requests took 129–131 ms after an 11.59 s cold request. A separate 62,431-position multimodal chat prompt generated 512 tokens at 71.2 output tokens/s. The [benchmark report](docs/BENCHMARKS.md#64k-context-and-vision-smoke-test) describes these operational probes and their limits.
+![Winnow-E4B and 12B, side by side: Q8 GPU memory, chat generation, and E4B cached 64K decision timing](assets/e4b-profiles.png)
 *Q8 operational profiles on an RTX 5070 Ti. The 8K text workload was matched; the 64K vision and chat probes used different images and question counts. The 10 GB+ and 12 GB+ capacity suggestions are inferred from measured memory, not tests on those cards.*

assets/e4b-profiles.png CHANGED Viewed

Git LFS Details

SHA256: fd32b0da47bd718ac7186aeb69f2582a94f7f620ee5f0b10e40ae2929018ba9f
Pointer size: 131 Bytes
Size of remote file: 261 kB

Git LFS Details

SHA256: f00e0b482eccdc7f01abccdf040da66057147b3ddb6a253a6bcb00daf03b7526
Pointer size: 131 Bytes
Size of remote file: 264 kB