Peacebell — ask me about World War II

Model card: wayneworkman2012/peacebell-v1-291M

Everything you type here and every reply the model gives is logged to a private dataset. There is no sign-in: every conversation is logged anonymously. The logs are reviewed by the model's author and may be used to train future versions of Peacebell. Do not enter personal information.

License: the model weights are Apache 2.0 — free to use, modify and redistribute, including commercially, under that license's terms. The model card states them in full, together with the Peacebell name rule.

Peacebell is a 291M-parameter GPT trained from scratch on a curated WWII corpus. Custom architecture, running on a free CPU Space: replies stream more slowly than on the ZeroGPU demo, but nothing here spends your ZeroGPU quota, so take your time. It is strongest on focused factual questions; asked about anything else it says so and offers a wartime subject instead. Answers come from the model's memory only — it's closed-book, so don't paste source text and ask about it (it can't read provided context). It can state false facts; verify anything important. Tip: keep temperature at 0 (the default) for the most reliable answers — higher values add errors and drift.