Model library / tventurella
Mr.Chatterbox-2.7B-Chat is a 2.7 billion-parameter Llama-style chat model trained on 19th-century British text to simulate a Victorian gentleman.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
The model is presented as a conversational version of Mr.Chatterbox-2.7B, fine-tuned on authentic Victorian material and intended for historical role-play rather than factual assistance. It uses act tokens to steer reply type and was evaluated on behavioural fit to period register, not on factual accuracy.
It runs with a 2,048-token context, weights in float32, and is released under a CC-BY-NC-SA-4.0 licence. The publisher tventurella provides default sampling settings (temperature 0.3, top-k 50, repetition penalty 1.3, up to 300 tokens). End tokens are 123 and 119; both must be specified as stop tokens during inference.
Specification
| Repository | tventurella/Mr.Chatterbox-2.7B-Chat |
| Publisher | tventurella |
| Published | 7 October 2026 |
| Architecture | LlamaForCausalLM |
| Model type | llama |
| Parameters | 2,705,820,160 |
| Layers | 32 |
| Hidden size | 2,560 |
| Attention heads | 20, 20 key/value heads |
| Context length | 2,048 tokens |
| Vocabulary | 65,664 tokens |
| Weight format | float32 |
| Licence | cc-by-nc-sa-4.0 |
| Base model | tventurella/Mr.Chatterbox-2.7B |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 5.0 GiB | Full precision as released |
| fp8 | 2.5 GiB | 8-bit, near-lossless on most models |
| int4 | 1.3 GiB | 4-bit, smallest footprint |
The cache costs 320 KiB per token at 16-bit, so the full 2,048-token context of one request needs about 0.6 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 1 | 9.0 GiB | the full 2,048 tokens |
| RTX 5090 32GB | 1 | 16.2 GiB | the full 2,048 tokens |
| L40S 48GB | 1 | 30.6 GiB | the full 2,048 tokens |
| A100 80GB | 1 | 59.4 GiB | the full 2,048 tokens |
| H100 80GB | 1 | 59.4 GiB | the full 2,048 tokens |
| RTX PRO 6000 96GB | 1 | 73.8 GiB | the full 2,048 tokens |
| H200 141GB | 1 | 114.3 GiB | the full 2,048 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve tventurella/Mr.Chatterbox-2.7B-Chat \ --max-model-len 2048 \ --gpu-memory-utilization 0.90
Questions
The weights take about 10 GiB at the released precision. That fits on one RTX 4090 24GB, which leaves roughly 9 GiB for the key-value cache.
The repository declares cc-by-nc-sa-4.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 2,048 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email hello@aiqre.com with the model name and roughly what volume you expect.
Also in the library
EdgeRock-4B is a 4.6 billion-parameter multimodal language model with constant-memory reasoning capabilities.
EdgeRock-4B is a 4.6 billion-parameter multimodal language model with constant-size memory for text and image reasoning.
Clio is a 3 billion-parameter stablelm model released by NovelAI for fast text generation.
open-1b-sft is a 1.6 billion-parameter, decoder-only English chat model fine-tuned with supervised data.
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.