---
license: apache-2.0
language:
- en
library_name: gguf
pipeline_tag: image-text-to-text
model_name: Ornith-1.0-35B-TQ3_4S
tags:
- gguf
- llama.cpp
- qwen
- qwen3.6
- qwen-vl
- multimodal
- vision
- ornith
- turboquant
- tq3_4s
base_model:
- deepreinforce-ai/Ornith-1.0-35B
model-index:
- name: Ornith-1.0-35B-TQ3_4S
  results: []
---

# Ornith-1.0-35B-TQ3_4S

![Ornith-1.0-35B-TQ3_4S](thumbnail.png)

`Ornith-1.0-35B-TQ3_4S` is a compact TurboQuant GGUF build of [deepreinforce-ai/Ornith-1.0-35B](https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B).

## Required Runtime

This model uses the custom `TQ3_4S` tensor type and requires [turbo-tan/llama.cpp-tq3](https://github.com/turbo-tan/llama.cpp-tq3).

> Stock `llama.cpp` builds without TurboQuant support cannot load this model.

This is the standard 35B model, not an MTP release. Do not add draft-MTP speculative-decoding flags.

`Ornith-1.0-35B` is a Qwen-VL **multimodal** model. Vision is optional — load the `mmproj` projector to enable image input (see [Vision (Multimodal)](#vision-multimodal)).

## Files

- [`Ornith-1.0-35B-TQ3_4S.gguf`](Ornith-1.0-35B-TQ3_4S.gguf) - main model, 13.00 GiB
- [`mmproj-f16.gguf`](mmproj-f16.gguf) - vision projector (mmproj), F16, 0.84 GiB
- [`thumbnail.png`](thumbnail.png) - model card banner
- [`benchmark.png`](benchmark.png) - benchmark comparison card
- [`benchmark_notes.md`](benchmark_notes.md) - compact benchmark provenance and comparison

## Build the Required Runtime

```bash
git clone https://github.com/turbo-tan/llama.cpp-tq3
cd llama.cpp-tq3

cmake -S . -B build \
  -DCMAKE_BUILD_TYPE=Release \
  -DGGML_CUDA=ON \
  -DGGML_CUDA_FA_ALL_QUANTS=OFF \
  -DGGML_CUDA_GRAPHS=ON

cmake --build build --target llama-server -j
```

For an RTX 3090, `-DCMAKE_CUDA_ARCHITECTURES=86` may be added explicitly. Use the architecture matching your GPU on other systems.

## Recommended Runtime

Validated on an NVIDIA GeForce RTX 3090 Founders Edition with 24 GiB VRAM:

```bash
./build/bin/llama-server \
  -m Ornith-1.0-35B-TQ3_4S.gguf \
  --alias Ornith-1.0-35B-TQ3_4S \
  --host 127.0.0.1 --port 8080 \
  -c 32768 -np 1 -ngl 99 -fa on \
  -ctk q8_0 -ctv tq3_0 \
  --reasoning off --jinja
```

Runtime notes:

- `-fa on` enables flash attention at runtime.
- The validated CUDA build uses `GGML_CUDA_FA_ALL_QUANTS=OFF`.
- `-ngl 99` fully offloads the model on supported GPUs. Avoid partial offload when comparing the published speed.
- Reduce context from `32768` if the available VRAM is lower than 24 GiB.

## Quick Smoke Test

```bash
curl -s http://127.0.0.1:8080/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"Ornith-1.0-35B-TQ3_4S","messages":[{"role":"user","content":"Write ONLY the word ok."}],"max_tokens":16,"temperature":0.0}'
```

Expected assistant content:

```text
ok
```

## Vision (Multimodal)

`Ornith-1.0-35B` is a Qwen-VL multimodal model. To enable image input, add the vision projector with `--mmproj`:

```bash
./build/bin/llama-server \
  -m Ornith-1.0-35B-TQ3_4S.gguf \
  --mmproj mmproj-f16.gguf \
  --alias Ornith-1.0-35B-TQ3_4S \
  --host 127.0.0.1 --port 8080 \
  -c 32768 -np 1 -ngl 99 -fa on \
  -ctk q8_0 -ctv tq3_0 \
  --reasoning off --jinja
```

Send an image using the OpenAI vision schema (`image_url` with a base64 data URI):

```bash
curl -s http://127.0.0.1:8080/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"Ornith-1.0-35B-TQ3_4S","messages":[{"role":"user","content":[
        {"type":"text","text":"Describe this image."},
        {"type":"image_url","image_url":{"url":"data:image/png;base64,<BASE64_IMAGE>"}}
      ]}],"max_tokens":256,"temperature":0.0}'
```

Vision notes:

- The `mmproj` is only needed for image input; text-only use does not require it.
- Qwen-VL grounding works best with at least 1024 image tokens. If accuracy is poor on detailed images, add `--image-min-tokens 1024`.

## Benchmark Summary

Local BenchLoop and Hard86 comparison on an RTX 3090 FE using the TurboTan runtime and the launch settings above:

![Benchmark comparison](benchmark.png)

| Metric | Result |
|---|---:|
| 35B field overall | 95.75 |
| Hard86 | 81.4% |
| EasyCode | 100.0% |
| Toolcall | 88.3% |
| Data extract | 86.5% |
| Instruct follow | 65.5% |
| Reason math | 73.3% |
| Generation speed | 146.3 tok/s |
| Size | 13.00 GiB |

The field score uses `0.85 * task_score + 0.15 * size_factor`, with size normalized to the smallest displayed 35B model. Hard86 is weighted at `2x`, EasyCode at `0.5x`, and the remaining benchmark categories at `1x`.


## Validation Notes

- Benchmark results are local measurements, not claims from the parent model repository.

## License

Use is subject to the [base model](https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B) license and the licenses of [turbo-tan/llama.cpp-tq3](https://github.com/turbo-tan/llama.cpp-tq3).