7 Small Specialist Models That Beat Giant LLMs at One Job

Picture a support pipeline built the lazy way. A giant model transcribes the call, reads the attached invoice, pulls out the names and amounts, decides which team gets the ticket, and reads the reply back to the customer. It works, and every step is billed at frontier prices.

Most of those steps are single jobs, and small specialist models are built for exactly that. In this article, you will meet seven open models, each under 10 billion parameters, that go head to head with much bigger systems on one task. For each, you get the job it wins, its size and license, the published proof, and code to run it locally. These task-specific models won't replace a giant large language model (LLM) everywhere, and the last section covers where they don't.

How These Seven Were Picked

Every model had to meet four rules: open weights, one clear job, published proof against something much bigger, and a footprint small enough for a single machine. "Giant" does not always mean "a chat LLM." Where the proof is against a bigger general model, I say so. Where it is against a closed API or a larger specialist, I say that instead.

The Seven at a Glance

#	Job	Model	Size	License	What it beats
1	Transcription	Parakeet TDT 0.6B v3	0.6B	CC-BY-4.0	Whisper large-v3, on accuracy and speed
2	Optical character recognition (OCR)	PaddleOCR-VL-1.6	0.9B	Apache 2.0	Gemini 3 Pro and GPT-5.2 on OmniDocBench v1.6
3	Entity extraction	GLiNER	50M to ~300M	Check per checkpoint	ChatGPT on zero-shot named entity recognition (NER)
4	Embeddings	EmbeddingGemma	308M	Gemma terms	Open embedding models under 500M
5	Routing	Arch-Router	1.5B	Katanemo license	GPT-4o and Claude-Sonnet-3.7 on routing
6	Decisions	Nemotron-Orchestrator-8B	8B	NVIDIA License (research only)	GPT-5 on Humanity's Last Exam
7	Speech output	Kokoro-82M	82M	Apache 2.0	Larger open text-to-speech (TTS) models in blind listening tests

Setting Up

from huggingface_hub import login

login()  # paste a Hugging Face access token when prompted

Run it: