Which models should we add next? Roleplay & creative models for 2026 — call for supply and feedback

[flyworker]

Hi everyone,

We're refreshing the catalog for roleplay and creative-writing traffic, and we'd like your input before we decide. Two asks:

1. Models we'd like to add — looking for providers

Based on public 2026 usage rankings (e.g. OpenRouter's roleplay collection), these are models consumers ask for that we'd like providers for:

• deepseek-ai/DeepSeek-V4-Flash (0423) and DeepSeek-V4.1-Flash
• XiaomiMiMo/MiMo-V2.6-Flash
• TheDrummer/Artemis-31B-v1.1 (Gemma 4 31B base)
• TheDrummer/Skyfall-31B-v4.2 (Magistral Small base)

If you can serve any of these, reply with:

• the model, and the precision/engine you'd run (e.g. AWQ-4 on vLLM)
• the context window you can offer
• roughly how many concurrent requests you can take

2. Older roleplay models

Some of our Llama-3.3-70B-based roleplay models predate the 2026 bases. Our plan is to list newer alternatives alongside them first, and retire the older ones only once the new ones are serving well. If you run one of the older models today, tell us whether you'd switch, and to what.

Other suggestions are welcome too — models your users ask for, or fine-tunes you think deserve a listing. We'll post a summary of decisions in this thread.

Thanks!
— Swan Inference team

[flyworker]

MiMo-V2.6-Flash-RL and Qwen3.8-Queen-27B will go on trial first