Interfaze: The Future of AI is built on Task-Specific Small Models

2026-02-04

AI Summary

This work proposes a hierarchical architecture to address the high computational cost of conventional large language models in complex multimodal tasks and their inefficiency in integrating external information with specialized perceptual capabilities. The framework centers on task-specific small models, augmented by modules for context construction, multimodal perception (including OCR, ASR, and chart understanding), headless browser control, and sandboxed code execution. A lightweight controller orchestrates these components and forwards only refined context to a user-specified large language model for final response generation. This paradigm substantially reduces reliance on monolithic large models while achieving strong performance on multiple benchmarks: 83.6% on MMLU-Pro, 81.3% on GPQA-Diamond, 57.8% on LiveCodeBench v5, and 90.0% on AIME-2025. Most tasks are efficiently handled by the small models and tool stack, enhancing overall system efficiency and scalability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

We present Interfaze, a system that treats modern LLM applications as a problem of building and acting over context, not just picking the right monolithic model. Instead of a single transformer, we combine (i) a stack of heterogeneous DNNs paired with small language models as perception modules for OCR involving complex PDFs, charts and diagrams, and multilingual ASR with (ii) a context-construction layer that crawls, indexes, and parses external sources (web pages, code, PDFs) into compact structured state, and (iii) an action layer that can browse, retrieve, execute code in a sandbox, and drive a headless browser for dynamic web pages. A thin controller sits on top of this stack and exposes a single, OpenAI-style endpoint: it decides which small models and actions to run and always forwards the distilled context to a user-selected LLM that produces the final response. On this architecture, Interfaze-Beta achieves 83.6% on MMLU-Pro, 91.4% on MMLU, 81.3% on GPQA-Diamond, 57.8% on LiveCodeBench v5, and 90.0% on AIME-2025, along with strong multimodal scores on MMMU (val) (77.3%), AI2D (91.5%), ChartQA (90.9%), and Common Voice v16 (90.8%). We show that most queries are handled primarily by the small-model and tool stack, with the large LLM operating only on distilled context, yielding competitive accuracy while shifting the bulk of computation away from the most expensive and monolithic models.

Problem

• task-specific small models
• multimodal perception
• context construction
• large language models
• computational efficiency

Innovation

• task-specific small models
• context construction
• heterogeneous DNNs
• modular LLM architecture
• distilled context