turintech/ qwen3.5-0.8b-artemis:q5_k_m

56 4 hours ago

Text-only Qwen3.5 0.8B GGUF files. Default: q5_k_m. TurinTech models, built with Artemis.

tools thinking 
ollama run turintech/qwen3.5-0.8b-artemis:q5_k_m

Details

4 hours ago

a61cf487ab5f · 569MB

model
qwen35
·
752M
·
Q5_K_M
license
Apache License Version 2.0, January 2004 http://www.apache.org/licenses/ TERMS AND CONDITIONS FOR US
params
{ "presence_penalty": 1.5, "temperature": 1, "top_k": 20, "top_p": 0.95 }

Readme

Qwen3.5 0.8B Artemis

GGUF versions of Qwen3.5 0.8B for text. The vision encoder and the multi-token prediction layer are not included. This is an independent conversion, not affiliated with or endorsed by Alibaba Cloud or the Qwen team.

TurinTech models, built with Artemis

Ollama

ollama run turintech/qwen3.5-0.8b-artemis:q5_k_m

The packages use Ollama’s Qwen3.5 renderer and tool-call parser, the same as Ollama’s own qwen3.5 package, and Qwen’s recommended sampling for text in thinking mode: temperature 1, top_k 20, top_p 0.95, presence penalty 1.5. The context length is not pinned; set it to fit the available memory.

latest selects q5_k_m.

Tag Hugging Face file Size
q5_k_m (default) Qwen3.5-0.8B-Artemis-Q5_K_M.gguf 0.569 GB
q4_k_m Qwen3.5-0.8B-Artemis-Q4_K_M.gguf 0.532 GB

Each tag uses the corresponding Hugging Face file, verified by SHA-256. The comparison measurements used llama.cpp with thinking off, not Ollama.

Measurements and limitations

On the RTX 5090 GSM8K sample, Q5_K_M answered 132⁄250 correctly and Q4_K_M 137⁄250, versus 145⁄250 for BF16. The large tool-calling gains on the earlier panel used tasks that also influenced file selection; the fresh CPU probe did not establish a general tool-use advantage.

Full measurements and test conditions.

Licence

Apache-2.0, as released by Qwen. See the Hugging Face repository for LICENSE and NOTICE.

These files keep improving.