Limite 1B Violetto VRAM requirements

Limite 1B Violetto has 1.0B parameters. With an 8K-token context and one request it needs about 1.31 GB of GPU memory at Q4_K_M, 1.73 GB at FP8 and 2.79 GB at FP16/BF16. The published weights take 1.93 GB (BF16). The smallest setup here that holds it at Q4_K_M with an 8K context is the RTX 4060 8GB; on one 24 GB RTX 3090 / 4090 it runs with its full 128K-token context.

1.31 GB at Q4_K_M, 8K context, one request

FP8
1.73 GB
FP16 / BF16
2.79 GB
Published weights
1.93 GB
Smallest setup, Q4_K_M
RTX 4060 8GB
Open Limite 1B Violetto in the calculator

Worked example: Limite 1B Violetto with 32K tokens

Inputs: Limite 1B Violetto · Q4_K_M weights · 32,768 tokens of context · one request · FP16 KV cache

Weights (1.0B at Q4_K_M)
597 MB
KV cache (12 KB per token)
447 MB
Buffers and runtime (0.5 GB + 10%)
616 MB
Total
1.62 GB

The smallest setup here that holds it is the RTX 4060 8GB, with about 6.38 GB to spare. Change the inputs in the calculator

What makes Limite 1B Violetto's memory use different

Its 36 sliding-window layers keep only the last 1,025 tokens (llama.cpp gives them 1,792 cells: the window plus a 512-token batch, rounded up to 256): at its 128K limit they hold 63 MB of cache instead of the 4.50 GB they would need with full attention.

How much VRAM does Limite 1B Violetto need?

Limite 1B Violetto needs 1.31 GB at Q4_K_M, 1.80 GB at Q8_0 and 2.79 GB at FP16 with 8,192 tokens of context and one request. Each total below is the weights plus the FP16 KV cache and the runtime overhead (0.5 GB plus 10%), and each row opens the calculator with that setting. What the GGUF names mean.

PrecisionWeightsTotalSmallest setup
As published (BF16) 1.93 GB 2.79 GB RTX 4060 8GB
FP16 / BF16 1.93 GB 2.79 GB RTX 4060 8GB
FP8 / INT8 987 MB 1.73 GB RTX 4060 8GB
INT4 (AWQ / GPTQ) 525 MB 1.23 GB RTX 4060 8GB
GGUF Q8_0 1.02 GB 1.80 GB RTX 4060 8GB
GGUF Q6_K 810 MB 1.54 GB RTX 4060 8GB
GGUF Q5_K_M 700 MB 1.42 GB RTX 4060 8GB
GGUF Q4_K_M 597 MB 1.31 GB RTX 4060 8GB
GGUF IQ4_XS 537 MB 1.25 GB RTX 4060 8GB
GGUF Q3_K_M 483 MB 1.19 GB RTX 4060 8GB
GGUF IQ3_XXS 407 MB 1.11 GB RTX 4060 8GB
GGUF Q2_K 413 MB 1.11 GB RTX 4060 8GB

Fine-tuning Limite 1B Violetto? Limite 1B Violetto VRAM for LoRA, QLoRA and full training.

Limite 1B Violetto at 8K, 32K and 128K (full) tokens of context

12 of its 48 layers use full attention and 36 keep a sliding window of 1,025 tokens. Each extra token of context adds 12 KB of FP16 cache per request once the sliding windows are full. The last column is the smallest setup here, one card or a group, that holds it at Q4_K_M. How the KV cache works.

ContextKV cache, FP16Total at Q4_K_MTotal at Q8_0Smallest setup, Q4_K_M
8K tokens 159 MB 1.31 GB 1.80 GB RTX 4060 8GB
32K tokens 447 MB 1.62 GB 2.11 GB RTX 4060 8GB
128K tokens 1.56 GB 2.86 GB 3.34 GB RTX 4060 8GB

How fast Limite 1B Violetto writes

Tokens per second for one request with 8,192 tokens of context, estimated from memory bandwidth. A dash means it does not fit on one card. Try other settings in the speed calculator.

HardwareBandwidthQ4_K_MFP8
RTX 3060 12GB 360 GB/s 182–291132–202
RTX 4090 1,008 GB/s 341–646273–478
RTX 5090 1,792 GB/s 434–917368–717
M4 Max Mac (128 GB) 546 GB/s 241–410182–291
M3 Ultra Mac Studio (512 GB) 819 GB/s 307–558240–407
H100 SXM 3,350 GB/s 518–1,226465–1,022
H200 4,800 GB/s 555–1,388511–1,199

Longest context on one GPU

How many tokens of context Limite 1B Violetto fits on a single card with one request, an FP16 KV cache and 0.5 GB left free. "Full" means the model's whole context window fits.

GPUMemoryQ4_K_MQ8_0FP8
RTX 3060 12 GB 128K (full)128K (full)128K (full)
RTX 4060 Ti 16GB 16 GB 128K (full)128K (full)128K (full)
RTX 3090 / 4090 24 GB 128K (full)128K (full)128K (full)
RTX 5090 32 GB 128K (full)128K (full)128K (full)
A100 40GB 40 GB 128K (full)128K (full)128K (full)
Mac, 64 GB unified memory 48 GB 128K (full)128K (full)128K (full)
L40S / RTX 6000 Ada 48 GB 128K (full)128K (full)128K (full)
A100 / H100 80GB 80 GB 128K (full)128K (full)128K (full)
Mac, 128 GB unified memory 96 GB 128K (full)128K (full)128K (full)
H200 141 GB 128K (full)128K (full)128K (full)
B200 180 GB 128K (full)128K (full)128K (full)

Model details

Parameters
1.0B (1,035,253,888)
Layers
12 of its 48 layers use full attention and 36 keep a sliding window of 1,025 tokens
Attention cache
2 KV heads × 128
Context length
131,072 tokens
Published weights
1.93 GB (BF16)
On Hugging Face
paradigma-inc/limite-1b-violetto

Why this estimate looks this way

At Q4_K_M and an 8K-token context, Limite 1B Violetto uses 597 MB for weights, 159 MB for its FP16 KV cache and 588 MB for estimated runtime overhead, totaling 1.31 GB. The overhead is a 0.5 GB base plus 10% of weights and cache.

Its attention layout matters for long context: 12 of its 48 layers use full attention and 36 keep a sliding window of 1,025 tokens. The FP16 cache grows by about 12 KB per additional token per request once the sliding windows are full.

The architecture and published checkpoint size come from the model's config.json and weight files. These are estimates rather than measured peak memory; inference engines can reserve extra buffers or preallocate the full configured cache.

Limite 1B Violetto next to similar-sized models

The three models here closest to it in parameter count, at Q4_K_M with 8,192 tokens of context.

ModelParametersTotalKV per tokenSmallest setup
Limite 1B Violetto 1.0B 1.31 GB 12 KB RTX 4060 8GB
MiniCPM5 2B 2.5B 2.42 GB 42 KB RTX 4060 8GB
Nemotron 3 Nano 4B 4.0B 3.10 GB 16 KB RTX 4060 8GB
Spark-X2.5 4B 4.1B 3.47 GB 36 KB RTX 4060 8GB

Badge for your model card

Paste it into a Hugging Face or GitHub README; it shows the Q4_K_M total above and links to this page. Q8_0 and FP16 badges.

Limite 1B Violetto VRAM: 1.31 GB at Q4_K_M, 8K context

Numbers read from the model files on Hugging Face on . Compare it with other models in the reproducible model and GPU report.