Models
The AI layer behind the infrastructure
Explore open-source and commercial models by type, deployment style, and practical operational fit across Black Scarab's industry focus areas.
Filters
Explore models by deployment reality
Model Type
Licensing
Deployment Style
Model Library
8 models matched
This section is meant to help evaluate the AI layer separately from the infrastructure layer, so model decisions stay clear instead of getting mixed into the hardware catalog.
Llama 3.1 8B
A compact general-purpose language model that can support local assistants, summarization, and workflow copilots on edge-capable hardware.
Best For
On-premise assistants, Private text processing, Workflow guidance and summarization
Black Scarab POV
This is one of the most practical starting points for organizations that want useful language capabilities without making the whole system dependent on external APIs.
Gemma 3 4B
A lightweight language model suitable for constrained edge systems, compact copilots, and local automation flows.
Best For
Small-footprint text applications, Embedded assistants, Fast local inference
Black Scarab POV
This kind of model matters when the hardware constraint is real. It can unlock usable AI on devices where larger models are impractical.
Qwen 2.5 7B Instruct
A versatile instruction-tuned model for multilingual workflows, operations copilots, and structured reasoning tasks.
Best For
Multilingual operations, Instruction-following workflows, Structured enterprise assistants
Black Scarab POV
This is attractive for LatAm-facing deployments because it helps bridge language flexibility with practical local deployment options.
Whisper
A speech recognition model for transcription, voice interfaces, and operational audio workflows.
Best For
Speech-to-text pipelines, Voice notes and field reporting, Call or interview transcription
Black Scarab POV
Speech is often underestimated in operational systems. Whisper becomes especially valuable when paired with field workflows that generate voice notes or operator reports.
YOLOv8
A real-time computer vision model family used for detection, tracking, and scene awareness in operational environments.
Best For
Object detection, Real-time video analytics, Monitoring and event detection
Black Scarab POV
This is one of the clearest examples of a model family that becomes more valuable when paired with the right camera, compute, and deployment discipline.
RT-DETR
A modern detection architecture built for high-quality vision tasks where precision and strong object understanding matter.
Best For
Higher-precision detection, Structured visual monitoring, Modern vision pipelines
Black Scarab POV
This is useful when the business problem needs better perception quality than lightweight real-time models alone can provide.
GPT-4o mini
A compact hosted multimodal model for text, image understanding, and workflow automation through an API-based architecture.
Best For
Cloud copilots, Multimodal workflow orchestration, Fast API-driven integrations
Black Scarab POV
A good fit when the business needs quick iteration and broad model capability, but we would not treat it as the primary brain for low-connectivity field systems.
Claude 3.5 Haiku
A hosted fast-response model for structured writing, summarization, customer workflows, and operational copilots.
Best For
Fast text processing, Operational copilots, Customer and support workflows
Black Scarab POV
This works well as a layer above operational systems, but it should complement edge infrastructure rather than replace it in field-critical environments.
Next Step
Pair the right model with the right stack
Once you know which model direction fits the use case, we can map it onto the right compute, sensing, and connectivity architecture.