Skip to content
VALEXII.AI

AUREX-OX · MODEL LABORATORY

RESEARCH CANDIDATES · UPDATED 2026-11-06

Global AI Model Research & Customization Program

Evaluating open and commercially adaptable AI technologies from leading international research ecosystems.

This is a research and customization candidate list. It is not a claim that these models have been downloaded, trained, fine-tuned or integrated. Only Qwen is an existing local experimentation baseline, and its live hosted connection remains unverified until demonstrated.

Evaluation status ladder / Evidence-gated

Statuses advance on evidence, never automatically.

Every candidate begins as a research candidate. Movement along the ladder requires verified work at each step.

  1. Research Candidate

  2. License Review

  3. Evaluation Planned

  4. Benchmarking

  5. Customization Trial

  6. Independently Verified

  7. Approved for Deployment

Model registry / Searchable

The current candidate list.

Initial experimentation priority: Qwen, Mistral/Ministral, K2 Horizon and Falcon-H1 — a strong first comparison before expanding into the remaining candidates.

12 of 12 candidates shown

  • Qwen

    Research Candidate

    China · Alibaba Qwen team · specific version not yet verified

    Existing local baseline, reasoning, coding, industrial document interpretation

    Customization method:
    Local experimentation baseline; RAG and parameter-efficient fine-tuning under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified
    Existing local experimentation baseline — live connection unverifiedInitial experimentation priority

  • Mistral / Ministral

    Research Candidate

    France · Mistral AI · specific version not yet verified

    Enterprise customization, reasoning, efficient inference

    Customization method:
    RAG and parameter-efficient fine-tuning (LoRA) under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified
    Initial experimentation priority

  • K2 Horizon

    Research Candidate

    United Arab Emirates · UAE research ecosystem · specific version not yet verified

    Reasoning, research evaluation, compact-model experimentation

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified
    Initial experimentation priority

  • Falcon-H1

    Research Candidate

    United Arab Emirates · Technology Innovation Institute · specific version not yet verified

    Efficient hybrid-model testing and adaptation

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified
    Initial experimentation priority

  • AIN / KITAB-Bench

    Research Candidate

    United Arab Emirates · UAE research ecosystem · specific version not yet verified

    Multimodal document understanding and OCR evaluation

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • DeepSeek

    Research Candidate

    China · DeepSeek · specific version not yet verified

    Advanced reasoning and model comparison

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • Apertus

    Research Candidate

    Switzerland · Swiss AI research initiative · specific version not yet verified

    Transparent and reproducible model research

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • Selected permissively licensed open models

    Research Candidate

    United States · Various open research groups · specific version not yet verified

    General-purpose and enterprise benchmark baselines

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • EXAONE

    Research Candidate

    South Korea · LG AI Research · specific version not yet verified

    Industrial reasoning and multilingual evaluation

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • Sakana AI research

    Research Candidate

    Japan · Sakana AI · specific version not yet verified

    Experimental orchestration and model optimization

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • Sarvam

    Research Candidate

    India · Sarvam AI · specific version not yet verified

    Multilingual and specialized language-model evaluation

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified

  • SEA-LION

    Research Candidate

    Singapore · AI Singapore · specific version not yet verified

    Regional multilingual AI research

    Customization method:
    Under consideration
    Licensing review:
    Not completed
    Hardware requirements:
    Not yet assessed
    Independent benchmarks:
    None independently verified
Back to the laboratory Engineering status Discuss enterprise customization