Model library / paradigma-inc
Limite 1B, Violetto is a 1-billion-parameter dense autoregressive transformer aimed at high-throughput mathematical reasoning.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Limite 1B, Violetto is Paradigma's first model, a dense autoregressive transformer built for solving difficult mathematical problems at high throughput. The publisher describes it as a lightweight, minimally instruction-tuned system that excels in single-turn reasoning and is intended for mathematical tasks rather than general-purpose assistant use.
The model has 1 billion parameters and supports sequences up to 131 072 tokens. It is distributed in bfloat16 format under an Apache-2.0 licence. No expert layers are present, and the architecture follows the LimiteForCausalLM design. Deployment requires Python 3.12, vLLM 0.26.0, PyTorch 2.11.0 with CUDA 13.0, and can run on a single GPU with tensor and pipeline parallel sizes set to 1.
Specification
| Repository | paradigma-inc/limite-1b-violetto |
| Publisher | paradigma-inc |
| Published | 21 September 2026 |
| Architecture | LimiteForCausalLM |
| Model type | limite |
| Parameters | 1,035,253,888 |
| Layers | 48 |
| Hidden size | 1,280 |
| Attention heads | 10, 2 key/value heads |
| Context length | 131,072 tokens |
| Vocabulary | 151,680 tokens |
| Weight format | bfloat16 |
| Licence | apache-2.0 |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 1.9 GiB | Full precision as released |
| fp8 | 1.0 GiB | 8-bit, near-lossless on most models |
| int4 | 0.5 GiB | 4-bit, smallest footprint |
The cache costs 48 KiB per token at 16-bit, so the full 131,072-token context of one request needs about 6.0 GiB. Concurrency multiplies that number, not the weights.
That figure is an upper bound: this model uses sliding-window attention with a 1,024-token window on some layers, so long requests cache less than this.
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 | 17.2 GiB | the full 131,072 tokens |
| RTX 5090 32GB | 1 | 24.4 GiB | the full 131,072 tokens |
| L40S 48GB | 1 | 38.8 GiB | the full 131,072 tokens |
| A100 80GB | 1 | 67.6 GiB | the full 131,072 tokens |
| H100 80GB | 1 | 67.6 GiB | the full 131,072 tokens |
| RTX PRO 6000 96GB | 1 | 82.0 GiB | the full 131,072 tokens |
| H200 141GB | 1 | 122.5 GiB | the full 131,072 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve paradigma-inc/limite-1b-violetto \ --max-model-len 131072 \ --gpu-memory-utilization 0.90
Questions
The weights take about 2 GiB at the released precision. That fits on one RTX 4090 24GB, which leaves roughly 17 GiB for the key-value cache.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 131,072 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 [email protected] with the model name and roughly what volume you expect.
Also in the library
open-1b-sft is a 1.6 billion-parameter, decoder-only English chat model fine-tuned with supervised data.
K2-Horizon-0.9B is a 1.1 billion-parameter dense decoder-only model with a 128 k token context window.
Nanbeige4.2-3B is a compact agentic language model designed for reasoning, tool use and personal-assistant tasks.
Manacá-1B is a 1.7 billion-parameter Llama-style decoder-only model pretrained for Brazilian Portuguese.
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.