FAQ
Straight answers about an early-stage research project.
The most common questions about model type, local inference, Echelon, evidence status, privacy, licensing and reporting.
Are the public Quantum pilots chat models?
No. quantum-1-pilot and quantum-1.6-pilot are experimental base-completion models. They are not instruction-tuned assistants and no chat template or conversational alignment is promised.
What is GGUF?
GGUF is a model-file format commonly used by llama.cpp-compatible local inference tools. A GGUF file makes local experimentation possible, but does not establish quality, safety, compatibility or reuse rights.
Can the pilots run locally?
Public F16 GGUF files and a llama.cpp reference command are available. Minimum RAM, throughput and client compatibility have not been formally measured by the project.
Are Echelon Base and Echelon Chat separate model families?
No. They are planned stages or variants within the quantum-1-echelon model line. No trained Echelon model is currently public.
Why are there no standardized benchmark scores?
No versioned standardized downstream-task benchmark report is public. The website does not invent or infer scores and keeps the gap visible until a suitable reproducible evaluation exists.
What do the evidence labels mean?
Published means a public artifact directly supports the statement. Partial evidence means a configuration, preflight or smoke artifact exists without proving full completion. The remaining labels identify absent, unmeasured, incomplete or maintainer-dependent information.
Does local inference guarantee privacy?
No. Local execution may keep prompts on a device, but privacy depends on the application, logs, operating system, model source and surrounding infrastructure.
Are the model weights open source?
The public pilot files are downloadable, but their Hugging Face repositories do not state an explicit model and tokenizer license in the reviewed revisions. The website therefore does not describe them as open source or freely reusable.
Is the research peer-reviewed?
No academic peer-review or DOI is claimed. Publications on this website are project research notes linked to public artifacts and explicit evidence boundaries.
How can I report a problem?
Use a public GitHub issue for non-sensitive documentation or behavior reports, the private security process for vulnerabilities, and email for privacy, rights or sensitive-material concerns.
DOCUMENTATION DIRECTORY
Continue through the evidence.
Every section keeps public evidence, configured targets and missing information visibly separate.