Median Total Time
13.67s
Median TTFT
1.14s
Median Prefill TPS
216.56
Median Gen TPS
9.73
Context Size
262144
Quantization
r64
Engine
vllm
Creation Method
LoRA Finetune
Model Type
Gemma31B
Chat Template
Gemma4
Reasoning
Yes
Vision
Yes
Parameters
31B
Added At
7/25/2026
language:

Equinox draws its name from the balance between extremes. Trained on a balanced blend of Wayfarer 2's unforgiving dark adventures and Hearthfire's quiet slice-of-life storytelling, Equinox is equally at home in perilous dungeons and candlelit conversations.
If you want to easily try this model, you can do so at https://aidungeon.com. Note that Equinox requires a subscription to use.
We plan to continue improving and open-sourcing similar models, so please share any and all feedback on how we can improve model behavior. Below we share more details on how Equinox was created.
Quantized GGUF weights can be downloaded here.
Equinox 31B was trained with two epochs of SFT (Supervised Fine-Tuning) on top of Gemma 4 31B Instruct, using a balanced dataset that combines two distinct creative directions:
Wayfarer 2 data — Dark, consequence-driven text adventures where choices carry weight, the world pushes back, and survival is never guaranteed. This is the grit and edge of Equinox.
Hearthfire data — Longform writing focused on slice-of-life, character depth, and emotional beats. Extended scenes where nothing "happens" but everything matters. This is the warmth and patience of Equinox.
Both datasets are primarily second-person present tense, supplemented with a smaller third-person dataset in the same narrative style.
The following settings are recommended as a baseline. Nothing stops you from experimenting with these, of course.
"temperature": 0.8,
"repetition_penalty": 1.05,
"min_p": 0.025
As a balanced model, Equinox occupies a middle ground by design. If you specifically want maximally grim, unforgiving storytelling, Harbinger or Wayfarer 2 may better suit your needs. For pure atmospheric slowburn, Hearthfire remains the specialist.
No reasoning datasets were included during training.
Gemma 4's native chat template was used, and thinking mode is suppressed by default.
<bos><|turn>system
You're a masterful storyteller and gamemaster. Write in second person present tense (You are), crafting vivid, engaging narratives with authority and confidence.<turn|>
<|turn>user
> You peer into the darkness.<turn|>
<|turn>model
<|channel>thought
<channel|>You have been eaten by a grue.
GAME OVER<turn|>
Thanks to Gryphe Padar for collaborating on this finetune with us!