Local LLMs

Anarlog

Open-source on-device AI notepad for meetings — local transcription, BYO API keys, notes saved as portable files. Formerly Hyprnote; canonical brand is now Anarlog.

• Free
• meeting-notes
• transcription
• open-source

Best for: Privacy-conscious professionals (lawyers, healthcare, researchers, journalists), Operators in regulated industries where cloud notetakers are non-starters

AUTOMATIC1111

Feature-rich Stable Diffusion WebUI with extensive model, extension, and parameter control.

• Free
• image-generation
• stable-diffusion
• local-inference

Best for: Advanced local Stable Diffusion workflows

CogView 4

THUDM text-to-image model family for high-quality generation in open research and local workflows.

• Free
• image-generation
• text-to-image
• open-weights

Best for: Developer workflows, Faceless content production

ComfyUI TTS

Node-based text-to-speech and voice workflow stack inside ComfyUI using custom audio nodes.

• Free
• text-to-speech
• voiceover
• narration

Best for: Local custom voiceover pipelines, Experimental multi-model TTS workflows

Command R+

Large instruction-tuned model oriented to advanced assistant and retrieval-heavy workflows.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Advanced local assistant deployments, Complex retrieval and planning workflows

Coqui TTS

Open-source toolkit for local text-to-speech and voice cloning workflows.

• Free
• text-to-speech
• voiceover
• local-inference

Best for: Advanced local text-to-speech pipelines

DeepSeek-R1

Reasoning-focused open-weight family with MIT core licensing and smaller distilled options.

• Free
• local-inference
• open-weights
• mit

Best for: Reasoning-heavy workflows on distilled checkpoints, Local experimentation with open model pipelines

DeepSeek-V4

Preview open-weight DeepSeek family with Pro and Flash MoE models, 1M context, and strong coding and agentic reasoning focus.

• Free
• cloud-llm
• local-inference
• open-weights

Best for: Coding-agent experiments with open-weight models, Long-context analysis over documents or repositories

DeepSeek-VL2

Mixture-of-experts local vision-language family for OCR, documents, charts, and grounded multimodal reasoning.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Private visual document analysis, Multimodal document understanding

EchoMimic

Open-source audio-driven portrait animation framework with editable landmark control and newer multimodal animation branches.

• Free
• avatar-video
• local-inference
• open-source

Best for: Open diffusion-based avatar animation experiments

FLUX

FLUX family for quality-first generation, fast local variants, and modern in-context image editing workflows.

• Free
• image-generation
• text-to-image
• image-editing

Best for: Thumbnail and visual concept generation, Fast style exploration for creator content

Fooocus

Beginner-friendly local Stable Diffusion UI focused on high-quality images with minimal setup.

• Free
• image-generation
• stable-diffusion
• local-inference

Best for: Fast local image generation with minimal setup

Forge

Performance-focused Stable Diffusion WebUI fork designed for practical local generation speed and compatibility.

• Free
• image-generation
• stable-diffusion
• local-inference

Best for: Faster local Stable Diffusion workflows in a linear WebUI

Gemma 2

Older Gemma family branch focused on efficient local text workloads in 2B, 9B, and 27B sizes.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Efficient local chat workloads, Summarization and long-form drafting

Gemma 3

Multimodal Gemma family with 128K context and broad local deployment options under Gemma terms.

• Free
• local-inference
• open-weights
• on-device

Best for: Local assistants with manageable compliance processes, Multimodal summarization and extraction

Gemma 3n

Device-first Gemma branch with multimodal support, long context, and efficient E2B/E4B variants.

• Free
• local-inference
• open-weights
• on-device

Best for: Multimodal local assistant workflows, Privacy-sensitive visual assistant tasks

Gemma 4

Newest Gemma family with Apache-2.0 licensing, multimodal input, 256K context, and sparse on-device variants.

• Free
• local-inference
• open-weights
• on-device

Best for: Multimodal local assistant workflows, Multimodal document understanding

GLM-4.5 Air

Open-weight GLM model variant for local reasoning, coding, and automation workflows.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Private local LLM workflows, Reasoning and coding support in automation tasks

GLM-4.7-Flash

Lightweight GLM 4.7 branch focused on fast coding, reasoning, and long-context generation.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Fast local coding assistants, Reasoning-heavy drafting with tighter latency budgets

gpt-oss-20b

Apache-2.0 open-weight text model with long context and practical local deployment targets.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Private drafting and extraction workflows, Batch automations with stable cost control

Hallo

Open-source portrait animation model for higher-fidelity talking-head generation from one image and driving audio.

• Free
• avatar-video
• local-inference
• open-source

Best for: High-fidelity portrait animation from one image and audio

HiDream

Open HiDream family for quality-focused generation and instruction-based image editing.

• Free
• image-generation
• text-to-image
• image-editing

Best for: Thumbnail and visual concept generation, Fast style exploration for creator content

HunyuanDiT

Tencent open-weights diffusion-transformer image model family for high-quality text-to-image generation.

• Free
• image-generation
• text-to-image
• open-weights

Best for: Developer workflows, Faceless content production

InternVL 3.5

Apache-2.0 multimodal family with many size options and a strong focus on reasoning, OCR, and agent-style visual tasks.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Multimodal internal analysis workflows, Builders experimenting with vision-language tasks

InvokeAI

Polished local-first generative image platform with strong workflow UX for Stable Diffusion users.

• Free
• image-generation
• stable-diffusion
• local-inference

Best for: Professional local image workflows with cleaner UX

Kandinsky 3

Open-weights text-to-image model family oriented to prompt-following and stylistic generation.

• Free
• image-generation
• text-to-image
• open-weights

Best for: Thumbnail and visual concept generation, Faceless content production

Kimi K

Earlier open-weight Kimi branch for long-context reasoning and local LLM experimentation.

• Free
• local-inference
• open-weights
• reasoning

Best for: Local long-context drafting and analysis, Builders comparing open-weight LLM stacks

Kimi K2.6

Latest open-weight Kimi model for long-horizon coding, agent swarms, multimodal execution, and large-context local experimentation.

• Free
• local-inference
• open-weights
• reasoning

Best for: Local agentic coding workflows, Multimodal local assistant builds

Kokoro TTS

Compact open-weight TTS model for local voice synthesis and experimentation.

• Free
• text-to-speech
• voiceover
• local-inference

Best for: Lightweight local text-to-speech experiments

Kolors

Open-weights text-to-image model family from Kwai for high-quality image synthesis workflows.

• Free
• image-generation
• text-to-image
• open-weights

Best for: Thumbnail and visual concept generation, Faceless content production

LatentSync

Open-source lip-sync framework for generating talking portrait videos from audio and face inputs.

• Free
• avatar-video
• local-inference
• open-source

Best for: Free local talking-head generation

LivePortrait

Open-source local portrait animation tool that turns a single image into a talking video.

• Free
• avatar-video
• local-inference
• open-source

Best for: Local avatar animation workflows

Llama 3.1

Open model family often used as a balanced local default for general chat, writing, and coding.

• Free
• local-inference
• open-weights
• self-hosted

Best for: General local chat and assistant workflows, Summarization and drafting tasks

Llama 3.2 Vision

Vision-capable Llama model for local image-plus-text understanding tasks.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Local image + text analysis workflows, Multimodal document understanding

Llama 3.3

Larger Llama generation aimed at high-quality local reasoning and assistant workflows.

• Free
• local-inference
• open-weights
• self-hosted

Best for: High-quality local assistant workflows, Reasoning-heavy long-form tasks

Llama 4

Open-weight multimodal family with massive context, but significant policy and license constraints.

• Free
• local-inference
• open-weights
• multimodal

Best for: Large multi-document summarization pipelines, Multimodal internal analysis workflows

LocalAI

Open-source local AI runtime with OpenAI-compatible APIs for self-hosted LLM and multimodal workloads.

• Free
• local-inference
• self-hosted
• open-source

Best for: Local model serving and testing, Private local LLM workflows

LocalForge

Open-source local app for running AI models and workflows on your own machine.

• Free
• local-inference
• open-source
• self-hosted

Best for: Local model serving and testing, Private local assistant workflows

MiniCPM-V 2.6

Efficient local VLM with strong OCR, multi-image, and video understanding in an 8B-class footprint.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Private visual document analysis, Multimodal local assistant workflows

Ministral 3 8B

Apache-2.0 open-weight 8B model tuned for efficient local use with very long context.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Long-document summarization and extraction, Private local assistant workflows

Mistral NeMo

Mid-size model line that balances general reasoning, coding support, and local deployability.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Balanced local assistant workloads, Coding and reasoning mixed tasks

Mistral Small 4

Open hybrid Mistral model that combines instruct, reasoning, coding, OCR, and transcription in one 256K-context family.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Multimodal local assistant workflows, Multimodal document understanding

Mixtral 8x22B

Mixture-of-experts model family offering strong quality with favorable active-parameter efficiency.

• Free
• local-inference
• open-weights
• self-hosted

Best for: High-end local inference setups, Long-context reasoning workflows

Molmo

Open vision-language family from AI2 focused on strong multimodal quality with Apache-2.0 licensing.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Multimodal document understanding, Private visual document analysis

MuseTalk

Open-source real-time lip-sync framework for talking avatar and portrait video workflows.

• Free
• avatar-video
• local-inference
• open-source

Best for: Free local talking-head generation

NVIDIA Nemotron

Open model family for agentic AI with reasoning-focused releases across edge, single-GPU, and multi-GPU tiers.

• Free
• open-weights
• reasoning
• agentic-ai

Best for: Agentic AI prototyping, Reasoning-heavy developer workflows

Ollama

Local LLM runtime for running open models on your own machine with simple CLI and API workflows.

• Free
• local-inference
• self-hosted
• offline

Best for: Local model serving and testing, Privacy-first AI workflows

Phi-3 Mini

Lightweight Phi model family for fast local inference on modest hardware.

• Free
• local-inference
• open-weights
• on-device

Best for: Low-latency local chat and coding help, Entry-level local LLM deployments

Phi-3.5 Mini Instruct

MIT-licensed small model with long context, optimized for practical local and on-device use.

• Free
• local-inference
• open-weights
• on-device

Best for: Private drafting and summarization on modest hardware, Lightweight offline content automation

Phi-3.5 Vision Instruct

Compact MIT-licensed multimodal model for local image, OCR, chart, and multi-image reasoning tasks.

• Free
• local-inference
• open-weights
• on-device

Best for: Multimodal document understanding, Private visual document analysis

Phi-4

Higher-capability Phi model for instruction-following and reasoning-heavy local tasks.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Reasoning-heavy local workflows, Structured instruction and planning tasks

Phi-4 Reasoning

Reasoning-tuned Phi-4 variant for complex chain-of-thought style local workloads.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Complex reasoning and analytical tasks, Local private inference with explicit step logic

Piper TTS

Fast local neural text-to-speech engine for offline voice generation.

• Free
• text-to-speech
• voiceover
• local-inference

Best for: Local private text-to-speech pipelines

PixArt-Σ

Open-weights text-to-image model line focused on efficient high-resolution generation.

• Free
• image-generation
• text-to-image
• open-weights

Best for: Thumbnail and visual concept generation, Faceless content production

Qwen2.5

Versatile multilingual open model family with strong long-form writing and instruction-following behavior.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Multilingual content generation, Long-form drafting and rewriting

Qwen2.5 VL

Multimodal Qwen model family for local vision-language workflows.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Multimodal local assistant workflows, Private visual document analysis

Qwen3 8B

Apache-2.0 open-weight 8B model with 128K context, local-first deployment, and optional cloud API access.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Private local writing and rewriting, Multilingual content transformation

Qwen3.5

Native multimodal Qwen family with sparse MoE scaling, strong agent behavior, and a flagship 397B total / 17B active open model.

• Free
• local-inference
• open-weights
• self-hosted

Best for: Multimodal local assistant workflows, Private visual document analysis

Qwen3.6

Qwen3.6 family covering the hosted Qwen3.6-Plus flagship and the first open-weight Qwen3.6-35B-A3B release.

• Free
• cloud-llm
• local-inference
• open-weights

Best for: Teams choosing between hosted and local Qwen generation, Agentic coding workflows

Qwen3.6-35B-A3B

First open-weight Qwen3.6 model: a 35B total / 3B active multimodal MoE focused on agentic coding and practical local use.

• Free
• local-inference
• open-weights
• apache-2-0

Best for: Local agentic coding workflows, Multimodal local assistant builds

Reflow Studio

Open-source local dubbing workstation combining RVC voice cloning, Wav2Lip lip sync, and face enhancement.

• Free
• lip-sync
• dubbing
• voice-cloning

Best for: Local avatar animation workflows, Developers building lip-sync and dubbing workflows

SadTalker

Open-source audio-driven talking-face generator for creating avatar-style clips from still portraits.

• Free
• avatar-video
• local-inference
• open-source

Best for: Free local talking-head generation

Stable Diffusion

Open model family for text-to-image generation, spanning v1.x, v2.x, SDXL, and SD3/SD3.5.

• Free
• image-generation
• design
• thumbnails

Best for: Faceless content production, Solopreneur operations

SwarmUI

Local-first Stable Diffusion UI focused on multi-model orchestration and scalable generation queues.

• Free
• image-generation
• stable-diffusion
• local-inference

Best for: Local Stable Diffusion workflows with larger job queues

VideoReTalking

Open-source talking-head editing stack for re-syncing, re-voicing, and expression-aware face video edits.

• Free
• avatar-video
• local-inference
• open-source

Best for: Talking-head repair and re-sync for existing video

Voicebox

Local-first open-source voice cloning studio powered by Qwen3-TTS.

• Free
• text-to-speech
• voice-cloning
• local-inference

Best for: Local custom voiceover pipelines, Advanced local text-to-speech pipelines

Wav2Lip

Open-source lip-sync model for syncing speech to an existing face video or portrait clip.

• Free
• avatar-video
• local-inference
• open-source

Best for: Fast local lip-sync for recorded face video

Z-Image

Z-Image text-to-image family for high-fidelity generation and fast iterative visual production.

• Free
• image-generation
• text-to-image
• image-editing

Best for: Thumbnail and visual concept generation, Fast style exploration for creator content