exllamav3 quantizations of sophosympatheia's Glistening-Gem-31B-v2.1

2.25bpw_h6 18.368 GiB
4.25bpw_h6 25.515 GiB (This model)

---

Glistening-Gem-31B-v2.1

A Creative Merge  ·  31B  ·  Apache 2.0

This is a merge of TheDrummer/Artemis-31B-v1, zerofata/G4-MeroMero-v2-31B, and llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic using stock google/gemma-4-31B-it as a base.

This merge improves upon sophosympatheia/Glistening-Gem-31B-v1.0 in creativity and prose, and avoids the issues that caused v2.0 to be a failed recipe.

Known Issues

This v2.1 release fixes the issues with v2.0 by using stock Gemma 4 instead of an exotic base model that made customizations to the final output layer of Gemma 4 that do not merge cleanly with other models. However, v2.1 still rarely produces some artifacts in a similar manner to v1, such as occasionally fusing two words together that should be separated by a space, or misspelling a word. These artifacts are quite rare in my testing and can be made even rarer by running more conservative sampler settings, like raising Min-P.

I believe the artifacting behavior comes from llmfan46/gemma-4-Ortenzya-The-Creative-Wordsmith-31B-it-uncensored-heretic in the recipe, which would explain it in v1 too. That ingredient helps shake up the word choices, but it also contributes to instability. The v2.1 recipe mitigates that impact better than v1 through strategic layer management, but I haven't perfected it yet. I think you'll find this model's outputs worth the trouble of having to occasionally fix a typo.

Sampler Tips

You can use the master import JSON in this repo (Glistening-Gem-v2-ST-Import.json) to deploy the sampler settings below. I recommend using these values as a starting point for your own experiments. It's not like the model falls apart if you deviate from these settings, but they should be a reliable starting point for most creative tasks.

A note on DRY + Adaptive-P: Some people recommend not mixing these, but in practice I haven't noticed any issues. Feel free to drop DRY entirely if you want. Adaptive-P is already designed to avoid repetitions.

Sampler Settings

Try these settings as a starting point.

Temp 0.8 Min-P 0.1 Adaptive-P Target 0.6 Adaptive-P Decay 0.75 DRY Mult. 0.8 DRY Base 1.8 DRY Allowed Length 6

Prompting Tips

You can download the Glistening-Gem-v2-ST-Import.json file from this repo and import it directly into SillyTavern to get system prompt, chat template, and sampler settings all in one go.

Donations

If you feel like saying thanks with a donation, I'm on Ko-Fi

Quantizations

Please see the sidebar of the model card where a link to quantizations can be found, or click here for the list of them.

License

Apache 2.0, inherited down from Gemma.

Merge Details

This is a merge of pre-trained language models created using mergekit.

Merge Method

This model was merged using the DELLA merge method using densenet/Gemma-4-31B-StyleTune-heretic-ara as a base.

See the model intro for more details and the mergekit_config.yaml file in the repo for the full recipe.