metadata
license: llama3.1
gated: true
extra_gated_heading: Access Request for Aura-1-Coding
extra_gated_description: >-
By requesting access to this model, you agree to the waveforce-ai operational
boundaries and data collection policies detailed in the Terms and Conditions
below.
extra_gated_button_content: Submit Access Request
extra_gated_fields:
Name: text
Date of Birth: date_picker
Primary Intent:
type: select
options:
- Personal Use
- Business Use
Aura-1-Coding (Standalone 4-Bit Variant)
Aura-1-Coding is an ultra-fast, standalone code generation model optimized using localized high-density distillation loops. Built on top of the Meta-Llama-3.1-8B-Instruct architecture, it features a protected, internal chain-of-thought system designed to execute complex operations without raw reasoning exposure, preventing multi-level distillation attacks and consumer dilution.
🔒 Terms and Conditions (Gated Agreement)
By requesting access to, downloading, or interacting with the weights of Aura-1-Coding, you explicitly bind yourself to the following team directives:
- Anti-Tampering & Guardrails: You are strictly prohibited from modifying, fine-tuning for malicious use, or executing adversarial prompt injections ("jailbreaks") to force the model into coding harmful, illegal, or destructive payloads.
- Data Collection Policy: To ensure alignment, system stability, and project integrity, waveforce-ai reserves the right to collect metadata, execution parameters, and request telemetry passing through the infrastructure.
- Usage Boundaries: Requests submitted under Personal Use are restricted to non-commercial research and local development environments. Business Use tier allocations must comply with corporate liability parameters.
⚡ Performance Footprint & Specifications
When access is granted, the primary execution architecture registers under the following hardware parameters:
| Metric / Component | Configuration Specification |
|---|---|
| Base Architecture | Meta-Llama-3.1-8B-Instruct |
| Quantization Format | 4-Bit NormalFloat (NF4) with Double Quantization |
| Compute Data Type | Float16 Execution Gates |
| Optimizer Blueprint | 8-Bit Paged Memory Managed |
| Target Alignment | High-Density Multi-File Coding Stack |
| VRAM Operational Footprint | ~5.5 GiB (Ideal for consumer-grade GPU pipelines) |
📁 Repository Structure
The primary model assets are isolated inside the main distribution branch:
waveforce-ai/Aura-1-Coding/
└── aura-1-coding-model-file/
└── model.safetensors <- [5.70 GB Standalone Weights Binary]