SAVRN
Search Contact SAVRN

Open-weight model · Text generation

NULLXES-SHINRA-4B-SMOKE

by Maga MagistrTheOne/NULLXES-SHINRA-4B-SMOKE

NULLXES SHINRA-4B-INSTRUCT is the Language Intelligence Layer of the NULLXES system. SHINRA is responsible for multilingual understanding, coding intelligence, instruction following, structured outputs, and agent preparation.

Parameters3.9B
Context32,768
Weights7.9 GB
Licenseother
AccessOpen weights
Monthly Downloads

Runs On

What it takes to serve NULLXES-SHINRA-4B-SMOKE (3.9B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

PrecisionWeightsMemory neededCheapest setupPer hourAlso fits
16-bit 7.9 GB 9.4 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
8-bit 3.9 GB 4.7 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00
4-bit 2.0 GB 2.4 GB 1x MI300X (192 GB)
Vultr
$1.85 1x H100 $1.99 · 1x MI325X $2.00

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Sep 18, 2026.

Model Card

NULLXES SHINRA-4B-INSTRUCT is the Language Intelligence Layer of the NULLXES system. SHINRA is responsible for multilingual understanding, coding intelligence, instruction following, structured outputs, and agent preparation. This checkpoint is the instruction-tuned (and optionally DPO-aligned) 4B-class dense decoder. Proprietary ShinraForCausalLM (not a Llama / Mistral / Qwen / GPT-NeoX wrapper). RMSNorm → GQA+RoPE → residual → RMSNorm → SwiGLU → residual then final RMSNorm and tied LM head. Special tokens:. Generation stop is. Document stop is. Three stages. Pretrain → NULLXES SHINRA-4B-BASE SHINRAPRETRAINV1: 40% FineWeb-Edu, 20% code (python-edu + licensed Stack), 15% math/science…

Excerpt from the card by Maga, licensed other.

Configuration

Architecture
ShinraForCausalLM
Context length (tokens)
32,768
Layers
32
Hidden size
3,072
Feed-forward size
9,216
Attention heads
24
Key/value heads
8
Head dimension
128
Vocabulary size
131,072
RoPE base
1e+06
Model type
nullxes_shinra

Identity and Version

Repository
MagistrTheOne/NULLXES-SHINRA-4B-SMOKE
Publisher
Maga
Task
Text generation
Modality
Text
Library
transformers
Parameters
3.9B parameters
Languages
en, ru, de, fr, es, zh, ja, ko
Revision
a2483c96fd5d50b00d4eb62f947b01df4ae73509
First published
2026-09-18
Last updated
2026-09-18

Files and Weights

13 files, 7.9 GB in total. The weights are 1 file totalling 7.9 GB in safetensors.

Weights1 file · 7.9 GB
Configuration7 files · 46.3 KB
Tokenizer2 files · 8.8 MB
Documentation1 file · 5.3 KB
Other1 file · 1.4 KB
Repository1 file · 1.5 KB
Every file
FileTypeSizeSHA-256
model.safetensorsWeights7.9 GB 3a54f0239ffe
config.jsonConfiguration1.2 KB
configuration_shinra.pyConfiguration7.2 KB
modeling_attn.pyConfiguration18.3 KB
modeling_mlp.pyConfiguration945 B
modeling_norm.pyConfiguration1.1 KB
modeling_shinra.pyConfiguration17.6 KB
train_state.jsonConfiguration35 B
README.mdDocumentation5.3 KB
chat_template.jinjaOther1.4 KB
.gitattributesRepository1.5 KB
tokenizer.jsonTokenizer8.8 MB
tokenizer_config.jsonTokenizer859 B

License and Download

License
other
Access
Open weights, no gate
Download size
7.9 GB
Download from Maga

Released by Maga through its official repository on Hugging Face.

Memory Requirements

PrecisionWeights in memory
As published7.9 GB
16-bit7.9 GB
8-bit3.9 GB
4-bit2.0 GB

Weights only, from the published parameter count; the key-value cache and runtime add to this.

Questions About NULLXES-SHINRA-4B-SMOKE

How much GPU memory does NULLXES-SHINRA-4B-SMOKE need?

About 9.4 GB at 16-bit and 2.4 GB at 4-bit: the weights (3.9B parameters) plus a working margin. A long context needs more.

What is the cheapest GPU to run NULLXES-SHINRA-4B-SMOKE on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

What license is NULLXES-SHINRA-4B-SMOKE released under?

other, as its publisher declares it. Read the license text before commercial use.

What is NULLXES-SHINRA-4B-SMOKE's context length?

32,768 tokens, from the maximum position embeddings in its published configuration.

Similar Models

Model · Text generation

NULLXES-SHINRA-4B-INSTRUCT

Maga

Language Intelligence layer of the NULLXES Intelligence Stack. SHINRA Our llm. Release line 1. NULLXES SHINRA-4B-BASE — pretrain 2. NULLXES SHINRA-4B-INSTRUCT — instruction tuning ← this model 3. NULLXES SHINRA-4B-INSTRUCT (aligned) — DPO / preference optimization Parameter budget Architecture source: configs/shinra4b.yaml at 903a639e03bc. The total also matches the published safetensors metadata. Custom SentencePiece Unigram, trained in-house on a web + wiki + code mix. Special tokens trustremotecode=True is required: SHINRA ships a custom ShinraConfig and modeling code, not a reused architecture class. The HF inference widget cannot load custom code, hence inference: false. Data pipeline…

Open weights other 3.9B parameters 32,768 tokens transformers

Model · Text generation

NVIDIA-Nemotron-3-Nano-4B-BF16

NVIDIA

Dec 2025 \- Jan 2026 September 2024 The pretraining data has a cutoff date of September 2024\. NVIDIA-Nemotron-3-Nano-4B-BF16 is a small language model (SLM) trained from scratch by NVIDIA, and designed as a unified model for both reasoning and non-reasoning tasks. It responds to user queries and tasks by first generating a reasoning trace and then concluding with a final response. The model's reasoning capabilities can be controlled via a system prompt. If the user prefers the model to provide its final answer without intermediate reasoning traces, it can be configured to do so, albeit with a slight decrease in accuracy for harder prompts that require reasoning. Conversely, allowing the…

Open weights other 4B parameters 262,144 tokens transformers

Model · Text generation

Qwen3-4B

Qwen

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features: - Uniquely support of seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue) within single model, ensuring optimal performance across various scenarios. - Significantly enhancement in its reasoning capabilities, surpassing previous QwQ (in thinking mode) and…

Open weights apache-2.0 4B parameters 40,960 tokens transformers

Model · Text generation

Qwen3-4B-Instruct-2507

Qwen

We introduce the updated version of the Qwen3-4B non-thinking mode, named Qwen3-4B-Instruct-2507, featuring the following key enhancements: - Significant improvements in general capabilities, including instruction following, logical reasoning, text comprehension, mathematics, science, coding and tool usage. - Substantial gains in long-tail knowledge coverage across multiple languages. - Markedly better alignment with user preferences in subjective and open-ended tasks, enabling more helpful responses and higher-quality text generation. - Enhanced capabilities in 256K long-context understanding. Qwen3-4B-Instruct-2507 has the following features: NOTE: This model supports only non-thinking…

Open weights apache-2.0 4B parameters 262,144 tokens transformers

Model · Text generation

Qwen3-4B-Base

Qwen

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Building upon extensive advancements in training data, model architecture, and optimization techniques, Qwen3 delivers the following key improvements over the previously released Qwen2.5: Qwen3-4B-Base has the following features: For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our blog, GitHub, and Documentation. The code of Qwen3 has been in the latest Hugging Face transformers and we advise you to use the latest version of transformers. With transformers<4.51.0, you will…

Open weights apache-2.0 4B parameters 32,768 tokens transformers

Model · Text generation

qwen3-4b-base-dapo-v4

Reliquary

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Building upon extensive advancements in training data, model architecture, and optimization techniques, Qwen3 delivers the following key improvements over the previously released Qwen2.5: Qwen3-4B-Base has the following features: For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our blog, GitHub, and Documentation. The code of Qwen3 has been in the latest Hugging Face transformers and we advise you to use the latest version of transformers. With transformers<4.51.0, you will…

Open weights apache-2.0 4B parameters 32,768 tokens transformers