Available models

Nemotron 3 Nano

nemotron-3-nano

NVIDIA’s compact agentic 4B: long-context assistants that stay on a laptop.

HY 2Bit

hy-2bit

A 1.8B model squeezed to a phone-app footprint: always-on, on-device.

Ministral 3

ministral-3

Mistral’s edge family with native vision: phones through workstations, fully local.

Qwen3 VL

qwen3-vl

See, read, and act on screen: images, video, OCR, and GUI agents on your machine.

Gemma 3

gemma-3

Gemini-derived open models with vision and 140+ languages, from phones to desktops.

Gemma 3N

gemma-3n

Google’s efficient open model, tuned to run locally on phones, tablets, and laptops.

TRANSLATE GEMMA

translate-gemma

Gemma 3 specialists for on-device translation across 55 languages.

MedGemma

medgemma

Medical text and images on-device, for teams that cannot send patient data to the cloud.

Gemma 4

gemma-4

Google’s open models for phones and IoT: audio, vision, and near-zero latency on your silicon.

Qwen3

qwen3

Think when you need it, answer fast when you don’t: coding, math, and 119 languages, local.

Qwen3.5

qwen3-5

A natively multimodal daily driver: text, vision, and compact reasoning on your machine.

GLM

glm

Agentic reasoning for software, browser automation, and multi-turn tools, locally.

Devstral 2

devstral-2

Mistral’s coding agents for repos, multi-file edits, and long engineering sessions on-device.

Nemotron Nano 3

nemotron-nano-3

NVIDIA hybrid MoE for agents: reasoning, tools, and 1M-token context on your hardware.

GPT-OSS

gpt-oss

OpenAI’s first open weights since GPT-2. Reasoning, agents, and tools you run at home.

Qwen3 Coder

qwen3-coder

A coding specialist for agentic engineering: function calling, tools, and repo-scale work locally.

DeepSeek R1

deepseek-r1

Open reasoning distillations that punch above their size: chain-of-thought without a closed API.

DeepSeek OCR

deepseek-ocr

Token-efficient OCR for scans, photos, and layouts that cannot leave the building.

DeepSeek Coder

deepseek-coder

A compact coding line for completion, explanation, and refactors on a laptop.

Llama 3.2

llama-3-2

Meta’s on-device Llama: lightweight 1B and 3B chat, with vision-class siblings.

Llama 3.1

llama-3-1

Meta’s 8B generalist for chat and writing: the classic Llama stack, fully local.

Tiny Aya

tiny-aya

Cohere’s compact multilingual family. 70+ languages, regional variants, no cloud round-trip.