Open models built for fast on-device inference — from mobile NPUs to microcontrollers.
Vision-language understanding for on-device multimodal applications.
Coming soon
Multi-token prediction for faster decoding on edge accelerators.
Coming soon
Structured function calling at the edge — natural language in, typed device intents out.
Coming soonBuild responsive, private conversations that run directly on the device. Opt.Gear keeps generation fast as chats grow longer, without sending every prompt to the cloud.
Turn natural-language requests into reliable actions. Opt.Gear can understand intent, select tools, and coordinate multi-step tasks with low-latency inference at the edge.
“How do I get to Marcus Building?”
The same hybrid design — grouped-query attention for global routing, the ConvKV-Gated Mixer everywhere else — scaled from a server GPU down to a microcontroller.
The strongest quality-per-parameter point in the family. Korean–English bilingual, long-document comprehension, reasoning-tuned SFT — with a decode path built to stay fast.
Low-latency deployment on constrained mobile and edge devices — same hybrid design, same 64K window as the 1B.
Natural language to structured board-control intents, inside 1 MB of weights on an ARM Cortex-M7.
On-device decoding is bound by memory traffic — it is what sets your tokens per second — and the KV-cache is what generates it: every new token adds another position that every later token has to read back. Gear replaces the local attention layers with a fixed-length convolutional state. That is where the speed comes from.
At a full 64K context a local layer holds 3 positions of state where global attention holds 65,536. The buffer is sized by the convolution kernel, so it is a property of the layer, not of the prompt.
Local mixing keeps a three-position buffer per layer, whether the context is 1K or 64K.
Dynamic QKV matmuls and softmax give way to static linear, depthwise convolution, and element-wise ops — no dynamic shapes in the local path.
A small set of GQA layers still carries full-window retrieval, so long-range recall survives the substitution.
Speed is the headline; this is the proof it didn't come at the model's expense. Gear-1B leads every open model at or below 1.2B parameters on all four Korean benchmarks — KMMLU, KoBEST, CLIcK and HAERAE — from 0.5T training tokens and no distillation. The only model ahead on Korean is Qwen3-1.7B: 70% larger, trained on 72× the data.
Gear-1B
Gemma-3-1B
Llama-3.2-1B
EXAONE4-1.2BGear-1BGemma-3-1BLlama-3.2-1BEXAONE4-1.2BGear-1BLlama-3.2-1BEXAONE4-1.2BGemma-3-1B
Qwen3-1.7BGear-1BLlama-3.2-1BGemma-3-1BEXAONE4-1.2BQwen3-1.7B| Benchmark | Gear-1B QAT |
+ RTN INT4 | Gemma-3-1B QAT |
+ RTN INT4 |
|---|---|---|---|---|
| English | ||||
| MMLU | 41.3 | 39.8 | 35.8 | 32.5 |
| HellaSwag | 54.0 | 53.9 | 51.4 | 48.5 |
| WinoGrande | 58.3 | 56.9 | 55.0 | 53.8 |
| Korean | ||||
| KMMLU | 35.0 | 33.8 | 26.9 | 21.2 |
| KoBEST | 60.4 | 59.8 | 56.2 | 55.1 |
| HAERAE | 42.7 | 39.9 | 31.2 | 23.6 |
Both models are trained quantization-aware, then the identical RTN INT4 post-training quantization is applied on top (technical report, Table 5). Gear-1B gives up 1.5 points of MMLU and 0.1 of HellaSwag; Gemma-3-1B drops 3.3 and 2.9 on the same pipeline — and 7.6 on HAERAE. A robust QAT foundation is what keeps PTQ from falling off a cliff.
| Benchmark | Gear-1B |
Gemma-3-1B |
Llama-3.2-1B |
EXAONE4-1.2B |
 LFM2.5-1.2B |
Qwen3-1.7B |
|---|---|---|---|---|---|---|
| Training tokens | 0.5T | 2T | 9T | 12T | 28T | 36T |
| Distilled† | no | yes | yes | yes | yes | yes |
| Korean | ||||||
| KMMLU | 36.0 | 30.7 | 29.9 | 32.6 | 29.2 | 41.6 |
| KoBEST | 60.3 | 59.5 | 51.8 | 50.7 | 59.1 | 62.8 |
| CLIcK | 39.7 | 37.8 | 30.6 | 32.9 | 38.3 | 49.4 |
| HAERAE | 44.2 | 35.3 | 32.6 | 30.2 | 33.9 | 52.0 |
| English | ||||||
| MMLU | 43.2 | 39.8 | 46.1 | 37.2 | 51.1 | 60.3 |
| GPQA | 30.3 | 23.2 | 28.8 | 25.3 | 27.3 | 24.8 |
| ARC-Easy | 70.8 | 70.0 | 72.4 | 58.3 | 79.2 | 80.0 |
| ARC-Challenge | 38.6 | 40.1 | 41.7 | 34.0 | 51.0 | 53.0 |
| HellaSwag | 56.0 | 60.0 | 61.1 | 42.4 | 62.1 | 60.3 |
| PIQA | 71.2 | 72.5 | 74.8 | 55.4 | 72.9 | 72.4 |
| WinoGrande | 58.2 | 58.7 | 62.1 | 53.9 | 60.0 | 60.9 |
| OBQA | 36.8 | 36.4 | 36.2 | 27.8 | 40.8 | 39.4 |
†Whether the reported model was trained against a larger teacher. Gear was not.
▲Highest score in the comparison set, at any parameter count.
*All figures from Language Model Evaluation Harness under one shared configuration, evaluated July 2026 — they may differ from numbers published elsewhere. Per-task few-shot settings are listed in the technical report.
Not a simulator, not a server extrapolation. Gear-1M runs generative inference on a $60 microcontroller board — and every number below was read off the device.
Post-training quantization costs small models the most — they have no spare capacity to absorb the rounding. Gear-1B-QAT is trained against INT4 weights from the start, so the deployed artifact is the thing that was optimized: 0.6 GB on disk, 98% of the FP16 Korean average.
| Korean benchmark | FP16 | PTQ INT4 | QAT INT4 |
|---|---|---|---|
| KMMLU | 36.0 | 33.2 | 35.5 |
| KoBEST | 60.3 | 56.8 | 59.7 |
| CLIcK | 39.7 | 36.4 | 39.1 |
| HAERAE | 44.2 | 40.9 | 43.0 |
| Average | 45.1 | 41.8 | 44.3 |
| Model | Precision | Device | Decode (tok/s) | Per token (ms) | Weights |
|---|---|---|---|---|---|
| Gear-1M | W4A32 | STM32H747I · Cortex-M7 @ 400 MHz | 20.0 | 50.0 | 0.9 MB |
| Gear-270M | INT8 | Snapdragon 8 Elite Gen5 · Hexagon NPU | 96 | 10.4 | 310 MB |
| Gear-270M | INT8 | iPhone 17 Pro (A19 Pro) · Neural Engine | 88 | 11.4 | 310 MB |
| Gear-1B | INT4 · QAT | Snapdragon 8 Elite Gen5 · Hexagon NPU | 41 | 24.4 | 620 MB |
| Gear-1B | INT4 · QAT | iPhone 17 Pro (A19 Pro) · Neural Engine | 37 | 27.0 | 620 MB |
Single-stream decode, greedy sampling. Preliminary — prefill throughput, long-context decode curves and full methodology ship with the technical report.
Getting the weights and putting them on a device are different moments. Both are one step: every model ships with its export artifacts already built.
# pip install transformers from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "OptAI/Gear-1B-Instruct" tok = AutoTokenizer.from_pretrained(model_id) model = AutoModelForCausalLM.from_pretrained(model_id)
Open weights in the format your stack already reads.
Prebuilt exports for the runtimes and accelerators we profile.
Prebuilt binaries for the accelerators we profile on — the conversion pipeline is our job, not yours.
Download the weights, evaluate the model, and bring fast inference to your device.
View the models ↗Language models, Opt.Gear included, can produce inaccurate content that does not represent OptAI's views. Review anything you intend to rely on, publish, or ship — especially for medical, legal, or financial use.