BEST FOR • CURATED

Best AI Tools for AI Open Source

Best for AI Open Source

We've curated 96 top AI tools specifically selected for ai open source use cases. Each tool is evaluated for quality, reliability, and unique capabilities that make it well-suited for ai open source workflows.

WHY THESE TOOLS

These tools are selected because they excel at ai open source. When choosing, consider:

  • How the tool's specific features align with your ai open source needs
  • Whether the tool offers the right balance of quality, speed, and cost for your use case
  • Integration capabilities if you need to incorporate into existing workflows
  • Scalability for your production requirements
RESULTS
96 tools • curated
Standalone agent-first platform with CLI, SDK, and managed agents
Added May 19, 2026
AI-powered IDE built as a fork of Visual Studio Code, designed with an 'agent-first' paradigm where autonomous AI agents plan, execute, and validate code
Why: Antigravity 2.0 is Google's most credible bid for the agentic IDE seat. The new CLI and SDK make it competitive with Cursor, Claude Code, and Codex for terminal-first and automation workflows.
Freemium Best for Google-Native Agents Visit
The LLM-Ready Web Scraper: Turn Websites into Markdown
Added Jan 31, 2026
Firecrawl is the industry-standard tool for turning entire websites into clean, LLM-ready markdown
Why: Firecrawl is the leader of the 'LLM-Data' movement. We picked it because it's the first scraper that actually understands what AI models need: clean, noise-free markdown without the overhead of traditional scraping libraries.
Enterprise Best for AI Data Extraction Visit
The Invisible OS: Pure Execution via Messaging
Added Jan 27, 2026
Moltbot (also known as Clawdbot) is the spearhead of the 'Invisible OS' movement, a shift away from fragmented apps and toward pure, autonomous execution via messaging
Why: Moltbot represents the death of the 'app for everything' era. We picked it because it's the first agentic assistant to prove that reasoning-based execution through simple chat is more powerful than manual task management in 10+ different apps.
Freemium Best for Agentic Automation Visit
The Open-Source Scraping Engine: High-Performance LLM Crawling
Added Jan 31, 2026
Crawl4AI is an open-source, high-performance web crawling and scraping engine specifically optimized for large language models
Why: Crawl4AI is the leading open-source alternative to proprietary scraping APIs. We picked it because it offers the most powerful 'local-first' crawling experience, giving developers full control over their data extraction pipeline without the per-page costs of cloud services.
Free Best for Open-Source Crawling Visit
The node graph the rest of the field is measured against
New Added Aug 8, 2026
ComfyUI is an open-source node-based interface for diffusion models
Why: If you want to know exactly what happened between the prompt and the image, this is the tool that shows you. Every step is a node you can open, change and re-run, and a workflow someone else built arrives as a file you can load rather than a screenshot you have to reverse engineer. That reproducibility is why it became the format the rest of the field builds around.
Free Best for control Visit
API access to thousands of models on Hugging Face
Added Feb 5, 2026
Provides API access to thousands of machine learning models hosted on the Hugging Face Hub
Why: Largest model repository with API access, making it the go-to platform for accessing diverse AI models.
Enterprise Best for Model Variety Visit
Fast inference platform for AI models
Added Feb 5, 2026
High-performance inference platform providing ultra-fast API access to large language models and other AI models
Why: Fastest inference platform available, making it ideal for real-time applications requiring low latency.
Enterprise Best for Speed Visit
Moonshot AI's 2.8-trillion-parameter open-weight flagship
New this month Added Jul 16, 2026
Kimi K3 is Moonshot AI's open-weight model released July 16, 2026, built at roughly 2
Why: Kimi K3 is one of the most credible open-weight challengers to closed frontier models this year, aggressive enough on pricing and scale that it moved markets, Fortune covered it as a 'DeepSeek shock' moment for AI stocks.
Freemium Best for Open-Weight Frontier Performance Visit
Open-source node canvas built around the edit, not the prompt
New Added Aug 8, 2026
Invoke is an open-source generative image platform combining a unified canvas with inpainting, outpainting and layer control, plus a node editor for building repeatable workflows
Why: Its centre of gravity is the canvas rather than the graph, which suits the way a lot of real work happens: generate something, then keep editing regions of it. The node editor is there when a job needs repeating, instead of being the only way in.
Freemium Best for iterative editing Visit
The Efficiency Revolution: Frontier Intelligence at 1/100th the Cost
Added Feb 5, 2026
DeepSeek is the architect of the 'DeepSeek movement,' a fundamental shift in AI development that prioritizes extreme efficiency over raw compute
Why: DeepSeek changed the game by proving that 'expensive' doesn't always mean 'better.' We picked it because it's the first model family to offer true frontier-level reasoning (R1), general intelligence (V3), and advanced vision/OCR (VL2) with an open-weight philosophy and an API price point that makes proprietary models look obsolete.
Freemium Best for Cost-Efficiency Visit
Meta's open-source large language model
Added Feb 5, 2026
Llama is Meta AI's open-source large language model family with multiple versions: Llama (February 2023), Llama 2 (July 2023), Llama 3 (April 2024), Llama 3
Why: Meta's flagship open-source LLM with strong performance, extensive model sizes, and permissive licensing for research and commercial use.
Free Best for Open Source Visit
European open-source and commercial LLM
Added Feb 5, 2026
Mistral AI provides high-performance large language models with both open-source and commercial offerings
Why: European LLM provider with strong open-source offerings, multilingual capabilities, and focus on data privacy and compliance.
Freemium Best for Europe Visit
Alibaba's multilingual open-source LLM
Added Feb 5, 2026
Qwen is Alibaba Cloud's family of large language models with multiple versions: Qwen-1
Why: Alibaba's high-performance multilingual LLM with strong Chinese language support, cost-efficient pricing, and comprehensive open-source availability.
Freemium Best for Multilingual Visit
Microsoft's efficient small language models
Added Feb 5, 2026
Microsoft Phi is a family of small, efficient language models designed for high performance with minimal parameters
Why: Microsoft's efficient small language models with strong reasoning capabilities, MIT licensing, and optimized for resource-constrained environments.
Free Best for Efficiency Visit
Open-Source Coding Agents for Private, Fine-Tuned Development
Added Feb 5, 2026
SERA is a family of open-source coding agents developed by the Allen Institute for AI (AI2)
Why: We added SERA because it is the leading open-source alternative for privacy-conscious developers. It empowers teams to build their own custom coding assistants that understand their specific architectural patterns.
Free Best for Developers Visit
Google's open-source lightweight LLM
Added Feb 5, 2026
Gemma is Google DeepMind's family of open-source large language models, serving as lightweight versions of Gemini
Why: Google's open-source LLM family with strong performance, permissive licensing, and specialized variants for vision and medical applications.
Free Best for Research Visit
The 'Next DeepSeek' Movement: o1-Level Reasoning at 1/100th the Cost
Added Jan 31, 2026
Kimi k1
Why: Kimi k1.5 is the first model to prove that o1-level reasoning is achievable through efficient, open-weight architectures. We selected it because it consistently matches or exceeds Claude 4.5 in technical benchmarks (AIME, MATH-500) while offering a 2M context window and a significantly lower API price point, making frontier intelligence accessible to everyone.
Freemium Best for Technical Reasoning Visit
The Open Vision-Reasoner: SOTA Multimodal Performance
Added Jan 31, 2026
Qwen 2
Why: We added Qwen 2.5-VL to the Open Frontier movement because it is currently the highest-performing open-weight vision model. It proves that open source can lead in multimodal reasoning, especially for tasks requiring high-resolution OCR and long-form video understanding.
Free Best for Open Vision Reasoning Visit
Databricks' high-performance open-source LLM
Added Feb 5, 2026
DBRX is a mixture-of-experts transformer model developed by Databricks and Mosaic ML
Why: Databricks' high-performance open-source LLM with strong benchmark results, efficient MoE architecture, and permissive licensing.
Enterprise Best for Performance Visit
Meta's Open Multimodal Standard
Added Jan 31, 2026
Llama 3
Why: We included Llama 3.2 Vision because it is the most widely supported open multimodal model in the world. Its integration into almost every AI tool and framework makes it the 'default' choice for open-weight vision reasoning.
Free Best for Open Ecosystem Support Visit
The Open Vision Frontier: 124B Multimodal Power
Added Jan 31, 2026
Pixtral Large is Mistral AI's flagship 124B parameter multimodal model, designed to compete directly with GPT-4o and Claude 3
Why: We added Pixtral Large because it represents the peak of European open-weight AI. It is one of the few open models that truly matches the visual reasoning depth of the top proprietary models, making it essential for the Open Frontier movement.
Freemium Best for Complex Visual Reasoning Visit
The Open-Source Vision Giant: 78B Multimodal Leader
Added Jan 31, 2026
InternVL 2
Why: We included InternVL 2.5 because it is a consistent leaderboard champion. It often outperforms much larger models in visual reasoning and OCR, making it a critical tool for developers who need GPT-4 level vision without the proprietary lock-in.
Free Best for Leaderboard-Topping Vision Visit
Open-source, model-agnostic terminal coding agent
Added Jul 7, 2026
OpenCode is an open-source terminal-based coding agent that connects to multiple language models and performs autonomous software engineering tasks from the command line
Why: OpenCode is the best 'bring-your-own-model' coding agent for developers who want full control. Because it is open source and runs in the terminal, it fits naturally into existing CI/CD and shell-centric workflows without locking you into a specific vendor.
Free Best for Terminal Coding Visit
DeepSeek's open-weight model with permanent pricing
Added May 31, 2026
DeepSeek V4-Pro is a high-performance language model from DeepSeek
Why: DeepSeek V4-Pro stands out for combining frontier-level performance with transparent, permanent pricing and open weights. It is a practical choice for teams that want to self-host or avoid unpredictable API costs.
Freemium Best for Predictable Pricing Visit
Open-source image-to-video with LoRA support
Added Feb 5, 2026
Generates high-quality videos with motion diversity from images using Wan 2
Why: Open-source + LoRA customization for advanced users who need fine-tuned control and self-hosting capabilities.
Free Best for Open Source Visit
Tencent's high-quality open video model
Added Feb 5, 2026
High-quality image-to-video generation from Tencent using open-source Hunyuan Video models
Why: Strong open-source option with good quality, making it ideal for self-hosting and customization workflows.
Free Best for Open Source Visit
Open-source 20B model with commercial-grade text rendering and advanced image editing
Added Jan 1, 2026
Generates high-quality images from text prompts using Alibaba's Tongyi Qianwen 20-billion parameter MMDiT model
Why: Top-performing open-source model with exceptional text rendering and advanced image editing capabilities, optimized for efficient deployment.
Free Best for Text Rendering Visit
The Open Image Standard: The Midjourney Killer
Added Jan 1, 2026
FLUX
Why: FLUX.2 represents the shift toward 'High-End Open Source.' We picked it because it matches Midjourney's aesthetic quality while offering the transparency and customizability that only an open-weight model can provide.
Freemium Best for Open-Weight Quality Visit
80B parameter open-weight coding powerhouse
Added Feb 6, 2026
Alibaba's latest open-weight model specialized for coding
Why: Qwen3-Coder-Next is the best 'Private Brain' for coders. Most AI tools send your secret code to the internet, but this one can live entirely on your own computer. It's just as smart as the big paid tools, but it keeps your work 100% private and safe.
Free Best for Open Coding Visit
The first research stack built entirely by AI agents
Added Feb 6, 2026
An open-source research stack spanning Python, JS, C++, and CUDA, engineered from the ground up by autonomous AI coding agents
Why: A glimpse into the future of engineering. It's the first major technical stack where the AI wasn't just a helper, but the lead architect and builder.
Free Best for AI Research Visit
The open-source Firebase alternative with Vector support
Added Feb 5, 2026
Supabase provides a unified backend stack including a Postgres database, authentication, and storage
Why: Supabase is the 'All-in-One Toolbox' for building AI apps. It gives you a database, a way for users to log in, and a place for the AI to store its memory all in one spot. It's the easiest way to go from an idea to a working app without needing 10 different services.
Enterprise Best for Backend Visit
The global infrastructure for open-source model deployment
Added Feb 5, 2026
Now integrated with Cloudflare's global network, Replicate allows you to run and fine-tune open-source models (Flux, Llama, Whisper) with a single API call and zero infrastructure management
Why: The 'GitHub' of model deployment. It democratizes access to the world's best open-source models with enterprise-grade scaling.
Enterprise Best for Open Source Visit
Meta's Segment Anything 3D for high-fidelity reconstruction
Added Feb 5, 2026
SAM3D v2 leverages Meta's latest Segment Anything technology to reconstruct 3D geometry from single or multiple images with extreme precision
Why: The most precise open-source 3D reconstruction tool. Its boundary awareness makes it unbeatable for complex object modeling.
Free Best for Research Visit
Alibaba's open-source MoE flagship with thinking modes
Added Apr 28, 2025
Qwen 3 is a 2025 open-weight Mixture-of-Experts model family from Alibaba Cloud, ranging from 0
Free Best for Open-Source Agents Visit
Alibaba's open coding-specialist model
Added Nov 12, 2024
Qwen 2
Free Best for Open Coding Visit
Open mathematical reasoning specialist
Added Aug 1, 2024
Qwen-Math is a family of open-weight models specialized for mathematical reasoning and problem solving, derived from Qwen 2
Free Best for Math Reasoning Visit
Earlier open-source Wan video generation model
Added Mar 1, 2025
Wan 2
Free Best for Open Video Generation Visit
Fast local FLUX.2 generation for personal hardware
Added Jun 1, 2025
FLUX
Free Best for Fast Local Generation Visit
Open-weight FLUX.2 for research and commercial use
Added Jun 1, 2025
FLUX
Free Best for Open Customization Visit
Open-source JSON-native text-to-image model built for controllable, enterprise-safe generation
Added Jun 26, 2025
BRIA FIBO is an 8B-parameter DiT text-to-image model trained on long structured JSON captions
Why: FIBO stands out for native JSON structured prompting and fully licensed training data, making it the strongest open-source choice for enterprises that need predictable, legally safe image generation.
Freemium Best for Controllable Image Generation Visit
Fast, lightweight FIBO pipeline designed for speed, efficiency, and on-prem deployment
Added Nov 11, 2025
BRIA FIBO Lite is a lightweight variant of the FIBO image generation pipeline
Why: FIBO Lite gives teams a FIBO-family option optimized for speed and data sovereignty, with a fully local deployment path that the full FIBO pipeline does not emphasize.
Freemium Best for Fast Local Image Generation Visit
High-volume DeepSeek inference with a 1M-token context window
Added Apr 24, 2026
DeepSeek V4-Flash is the efficient sibling of V4-Pro, offering a 1M-token context window and configurable thinking modes at a fraction of the API cost
Why: V4-Flash delivers the same 1M context and thinking modes as V4-Pro at roughly one-third the API cost, making it the practical default for most production workloads.
Freemium Best for High-Volume APIs Visit
The open-weight reasoning model that sparked the efficiency revolution
Added Jan 20, 2025
DeepSeek R1 is a 671B-parameter open-weight reasoning model that matches o1-class performance on math, code, and logic benchmarks through reinforcement learning on verifiable tasks
Why: R1 proved that open-weight models can match proprietary reasoning systems at a fraction of the cost, making it a landmark for reproducible AI research.
Freemium Best for Open Reasoning Visit
The 128K-context MoE flagship that introduced sparse attention
Added Dec 1, 2025
DeepSeek V3
Why: V3.2 introduced DeepSeek Sparse Attention and unified thinking modes, making it the architectural bridge that enabled the later 1M-context V4 family.
Freemium Best for Long-Context MoE Visit
744B-parameter open-weight MoE flagship for agentic planning and execution
Added Feb 11, 2026
GLM-5 is Zhipu AI's (Z
Why: GLM-5 anchors the open-weight GLM line as a commercially-usable Chinese flagship with a permissive license and a strong reasoning profile.
Paid Best for Open-Weight Frontier Visit
MIT-licensed MoE flagship for 8-hour autonomous coding sessions
Added Apr 7, 2026
GLM-5
Why: GLM-5.1 is the open-weight coding release that made Z.ai competitive on long-horizon agentic work while remaining MIT-licensed for unrestricted commercial use.
Paid Best for Long-Horizon Coding Visit
Google's open multimodal model for research and developers
Added Mar 12, 2025
Gemma 3 is an open-weights family of multimodal models from Google, ranging from 1B to 27B parameters
Free Best for Open Multimodal Visit
Tencent's 389B-parameter open-source MoE language model
Added Nov 4, 2024
Open-source Transformer-based Mixture-of-Experts language model with 389 billion total parameters and 52 billion active parameters
Why: Largest open-source Transformer-based MoE model from Tencent, ideal for researchers and builders who want to self-host a capable long-context LLM.
Free Best for Open-Source LLM Workloads Visit
Tencent's fast, cost-efficient flagship Hunyuan model
Added Jan 10, 2026
A 200K-context open-weight Hunyuan model optimized for speed while maintaining strong performance on general chat, coding, and agentic tasks
Why: Speed-optimized Hunyuan flagship with a 200K context window and strong price/performance for production APIs.
Freemium Best for Speed Visit
Tencent's general-purpose instruction-tuned Hunyuan 2.0 model
Added Jun 20, 2025
Open-weight instruction-tuned variant of Tencent's Hunyuan 2
Why: Versatile instruction-tuned Hunyuan model balancing capability and context for a wide range of tasks.
Freemium Best for General-Purpose Chat Visit
The deep-thinking variant of Hunyuan 2.0
Added Sep 15, 2025
Open-weight reasoning variant of Hunyuan 2
Why: Hunyuan 2.0's reasoning mode for tasks that benefit from longer thought chains.
Freemium Best for Reasoning Visit
Tencent's efficient small-scale MoE instruct model
Added Nov 15, 2025
A compact open-weight MoE instruction model with a 131K context window, listed as a cost-efficient everyday Hunyuan option via OpenRouter and Tencent Cloud
Why: Smallest listed Hunyuan instruct model, making it attractive for budget-conscious long-context deployments.
Freemium Best for Cost-Efficient Inference Visit
Tencent's latest open-source MoE flagship with tool use
Added Apr 22, 2026
Open-weight preview of Hunyuan 3 (Hy3), a 295B-parameter MoE model with 21B active parameters and a 262K context window
Why: Tencent's strongest open-source Hunyuan model to date, with competitive coding and agentic benchmarks.
Freemium Best for Coding and Agents Visit
Tencent's open-source bilingual text-to-image diffusion transformer
Added May 14, 2024
Open-source text-to-image diffusion transformer with fine-grained Chinese and English understanding, multi-turn prompt refinement, ControlNet, LoRA, and IP-Adapter support
Why: Leading open-source bilingual text-to-image model with strong Chinese prompt understanding and a rich ecosystem.
Free Best for Chinese Text-to-Image Visit
Tencent's 80B-parameter open-source MoE image generator
Added Sep 28, 2025
Native multimodal MoE image generation model with 80 billion total parameters and 13 billion active parameters
Why: The largest open-source image generation model, combining high parameter counts with efficient MoE inference.
Free Best for High-Resolution Image Generation Visit
Tencent's open-source immersive 3D world generator
Added Jul 26, 2025
Generates explorable, interactive 3D worlds from text prompts or images using panoramic proxies and semantic layering
Why: Rare open-source pipeline for generating explorable 3D worlds from text or images, useful for games and immersive media.
Free Best for 3D Worlds Visit
Tencent's latest open-source high-fidelity 3D asset generator
Added Jun 13, 2025
Open-source system for generating high-resolution textured 3D assets from text or images
Why: Current open-source Hunyuan 3D pipeline with professional texture, PBR, and Blender integration.
Free Best for Production 3D Assets Visit
Moonshot's open-weight multimodal generalist with agent swarms
Added Jan 27, 2026
Kimi K2
Why: Kimi K2.5 was Moonshot's first widely available open-weight multimodal generalist and remains a notable reference point for the K2 family before K2.6 and K3 arrived.
Freemium Best for Open Multimodal Agents Visit
Moonshot's open-weight multimodal successor with long-context coding stability
Added Apr 21, 2026
Kimi K2
Why: Kimi K2.6 improves on K2.5 with stronger long-context coding and is a practical open-weight alternative for teams that want multimodal agents without the cost of closed frontier models.
Freemium Best for Long-Context Coding Visit
Meta's open-weight flagship with native multimodal reasoning
Added Apr 5, 2025
Llama 4 Maverick is Meta's flagship open-weight model, released in April 2025 as part of the Llama 4 family
Why: Maverick is the top open-weight model Meta actually ships today, with strong multimodal reasoning and a practical API ecosystem, making it the default choice for open Llama deployments.
Free Best for Open Multimodal Reasoning Visit
Long-context, efficient open multimodal model for edge and single-GPU use
Added Apr 5, 2025
Llama 4 Scout is Meta's efficient Llama 4 variant, released in April 2025
Why: Scout is notable for its extreme 10M-token context window and efficient single-GPU deployment, making it the standout open model for very long documents and memory-heavy applications.
Free Best for Long Context Visit
Efficient 70B open model matching 405B quality
Added Dec 6, 2024
Llama 3
Why: Llama 3.3 is the practical sweet spot in the Llama family: it gives users near-frontier open-model quality in a 70B package that is far cheaper to host and fine-tune than the 405B model.
Free Best for Efficient Open LLMs Visit
The first frontier-scale open-weight language model
Added Jul 23, 2024
Llama 3
Why: Llama 3.1 405B remains a landmark open release: it proved open weights could compete with proprietary frontier models and still serves as a high-quality baseline for research and synthetic-data generation.
Free Best for Frontier Open Research Visit
Mistral's flagship open-weight multimodal frontier model
Added Apr 15, 2026
A 675B-parameter sparse mixture-of-experts model with 41B active parameters and a 262K context window, released under Apache 2
Why: Mistral Large 3 is one of the most capable permissive open-weight models available, offering frontier performance with the deployment flexibility of Apache 2.0 licensing.
Freemium Best for Open-Weight Frontier Visit
Unified open-source small model for chat, reasoning, vision, and coding
Added May 1, 2026
A 119B-parameter MoE model with 6B active parameters and a 256K context window, released under Apache 2
Why: Small 4 packs flagship-class reasoning, vision, and coding into a single open-source model that is efficient enough for high-throughput and local deployments.
Freemium Best for Efficient Open Multimodal Visit
Mistral's edge family of small, dense open-source models
Added Apr 15, 2026
A family of 3B, 8B, and 14B parameter dense models released under Apache 2
Why: Ministral 3 brings Mistral's open-weight lineage to edge devices, offering a strong 14B reasoning option and smaller variants for local and on-device use.
Freemium Best for Edge Deployment Visit
Mistral's open-weight speech understanding and TTS models
Added Jul 1, 2025
A family of open-weight speech models including a 24B production variant and a 3B edge variant, released under Apache 2
Why: Voxtral offers open-weight speech understanding and synthesis at a fraction of the cost of proprietary alternatives, making it practical for production voice agents.
Freemium Best for Voice AI Visit
Compact 30B open-weight model with configurable reasoning for agents
Added Jun 4, 2026
A 30B total / 3B active parameter hybrid Mamba-2 + Transformer MoE language model built for efficient on-device and edge agentic tasks
Why: The smallest open-weight member of the Nemotron 3 family, giving teams frontier-style reasoning and tool-use without data-center hardware.
Free Best for Efficient Agents Visit
120B open-weight hybrid MoE for efficient multi-agent reasoning
Added Jun 4, 2026
A 120B total / 12B active parameter hybrid Mamba-Transformer MoE language model with LatentMoE, multi-token prediction, and native NVFP4 pretraining
Why: Fills the gap between Nano and Ultra with a strong efficiency-to-accuracy ratio for agentic orchestration and latency-sensitive serving.
Free Best for Multi-Agent Efficiency Visit
Efficient 8B physical-AI omni-model for workstations
Added Jun 1, 2026
An 8B-class physical-AI omni-model (8B reasoner + 8B generator) optimized for efficient inference on workstation-grade NVIDIA hardware such as the RTX PRO 6000
Why: Brings Cosmos 3 physical-AI capabilities to smaller hardware so labs and individual developers can experiment without a data center.
Free Best for Workstation Physical AI Visit
4B physical-AI omni-model for real-time edge robotics
Added Jun 1, 2026
A 4B-class physical-AI omni-model (2B reasoner + 2B generator) optimized for real-time robotic policy and visual reasoning at the edge
Why: The smallest Cosmos 3 variant, built for real-time robotic perception and policy where latency and power matter most.
Free Best for Edge Robotics Visit
NVIDIA-aligned 253B Llama 3.1 for helpfulness and instruction following
Added Dec 1, 2024
A 253B-parameter variant of Llama 3
Why: NVIDIA's largest aligned Llama collaboration, offering a strong open-weight alternative for teams already standardizing on Llama architectures.
Free Best for Aligned Llama Performance Visit
NVIDIA-aligned 49B Llama 3.1 for balanced performance
Added Dec 1, 2024
A 49B-parameter variant of Llama 3
Why: A mid-size aligned Llama model that balances capability and deployment cost for teams using NVIDIA tooling.
Free Best for Balanced Llama Deployment Visit
NVIDIA-aligned 8B Llama 3.1 for efficient inference
Added Dec 1, 2024
An 8B-parameter variant of Llama 3
Why: A compact, NVIDIA-aligned Llama model for teams that need HelpSteer-tuned instruction following on limited hardware.
Free Best for Efficient Aligned Llama Visit
Open foundation model for humanoid robot reasoning and control
Added Jun 4, 2026
NVIDIA's open foundation model for humanoid robot reasoning and control, combining an Eagle-based vision-language backbone with a diffusion transformer (DiT) action head for language-conditioned manip...
Why: NVIDIA's open contribution to humanoid robot foundation models, enabling language-conditioned manipulation across diverse robot embodiments.
Free Best for Humanoid Robotics Visit
The foundational open-source text-to-image model
Added Oct 20, 2022
Stable Diffusion 1
Free Best for Foundation Ecosystem Visit
High-resolution open-source image generation
Added Jul 26, 2023
Stable Diffusion XL (SDXL) is a 2023 open-source text-to-image model that generates higher-quality, higher-resolution images than SD 1
Free Best for High-Resolution Open Images Visit
Fast one-step SDXL for real-time generation
Added Nov 28, 2023
Stable Diffusion XL Turbo is a distilled, fast variant of SDXL that can generate images in a single step or a few steps
Free Best for Fast Open Images Visit
Efficient SD3 variant for consumer hardware
Added Jun 12, 2024
Stable Diffusion 3 Medium is a 2B-parameter version of SD3 designed to run well on consumer GPUs
Free Best for Local SD3 Visit
Stability AI's largest 3.5 model with best quality
Added Oct 22, 2024
Stable Diffusion 3
Free Best for SD3.5 Quality Visit
Compact open-source image-to-3D model from Microsoft
Added Dec 1, 2024
TRELLIS Mini is a smaller, faster variant of the TRELLIS family from Microsoft Research, designed for efficient image-to-3D generation on limited hardware while preserving the core architecture's qual...
Free Best for Fast Local 3D Visit
High-quality open-source image-to-3D from Microsoft
Added Dec 1, 2024
TRELLIS Large is the larger, higher-quality variant of the TRELLIS family, producing detailed 3D assets from single images with better geometry and texture fidelity than the smaller variants
Free Best for High-Quality 3D Visit
Fast open-source image-to-3D from Stability AI and Tripo
Added Mar 5, 2024
TripoSR is an open-source, feed-forward image-to-3D model developed by Stability AI and Tripo AI
Free Best for Fast Open 3D Visit
Open-source image generation with flexibility
Added Feb 5, 2026
Generates images from text with open-source flexibility and community support using Stable Diffusion 3
Why: Open-source standard with extensive customization options, making it the foundation for many custom image generation workflows.
Free Best for Open Source Visit
Microsoft Research's open image-to-3D model
Added Jul 7, 2026
TRELLIS 2 is an open-source image-to-3D generation model from Microsoft Research, released in 2026
Why: TRELLIS 2 is a valuable open research model for image-to-3D. It is ideal for academics, indie developers, and anyone who wants to run 3D generation locally or build on top of open weights.
Free Best for Open 3D Research Visit
FLUX image model family (provider site)
Added Feb 5, 2026
Publishes the FLUX family of state-of-the-art image generation models including FLUX
Why: Important modern image model family to know and track, representing the cutting edge of open-source image generation.
Best for Images Visit
Open physical-AI omnimodel for robotics and AV
Added Jul 7, 2026
NVIDIA Cosmos 3 is an open physical-AI omnimodel released around May 31 to June 1, 2026
Why: Cosmos 3 is a major open contribution to physical AI. By simulating realistic worlds, it can accelerate training for robots and self-driving cars while reducing the need for dangerous or costly real-world trials.
Free Best for Physical AI Simulation Visit
OpenAI's conditional 3D model generation
Added Feb 5, 2026
Generates 3D objects from text prompts or images using OpenAI's Shap-E model, a conditional generative model for 3D assets
Why: OpenAI's open-source 3D generation model with comprehensive documentation and active community, representing state-of-the-art conditional 3D asset generation from text and images.
Free Best for Research Visit
Open-source MoE LLM with strong Chinese NLP and multimodal capabilities
Added Jan 1, 2026
Baidu ERNIE 4
Why: Leading Chinese LLM with strong multilingual capabilities, open-source availability, and cost-efficient MoE architecture.
Freemium Best for Chinese Visit
Advanced multilingual LLM with enhanced reasoning and long-context support
Added Jan 1, 2026
GLM-4
Why: Advanced Chinese LLM with strong multilingual capabilities, efficient inference, and comprehensive deployment options.
Freemium Best for Multilingual Visit
Open-source text-to-3D motion model with 200+ motion categories and production-ready exports
Added Jan 1, 2026
Hymotion 1
Why: Tencent's cutting-edge open-source text-to-3D motion model with production-ready output and extensive motion category support.
Free Best for 3D Motion Visit
NVIDIA's 550B open-weights reasoning model, built for inference speed
New Added Aug 4, 2026
Nemotron 3 Ultra is the largest member of NVIDIA's Nemotron 3 family, released 4 June 2026 at Computex
Why: The architecture is optimised for throughput rather than peak benchmark score, and it shows: over 400 output tokens per second at 550B parameters. It is also the most openly documented release at this scale, because publishing the datasets and post-training recipes lets you actually reproduce and extend the model instead of just running it.
Free Best for Open-Weight Throughput Visit
Multimodal model generating image, video and audio from one set of weights
New Added Aug 4, 2026
FLUX 3 is Black Forest Labs' multimodal foundation model, announced 23 July 2026
Why: The first credible attempt to collapse image, video and audio generation into a single model rather than a pipeline of separate ones, from the team behind the most widely self-hosted open image models. Access is the catch: video is gated early-access and the open-weight release has not shipped, so treat availability as limited until FLUX 3 Dev lands.
Paid Best Multimodal Generation Visit
Alibaba's 2.4-trillion-parameter flagship, currently in preview
New this month Added Jul 19, 2026
Qwen 3
Why: Qwen 3.8-Max is Alibaba's answer to the current wave of massive open-weight-adjacent models from Chinese labs, and the discounted preview pricing makes it worth evaluating early even before the full release details land.
Paid Best for Long-Horizon Agentic Work (Preview) Visit
753B open-weight MoE coding model with a 1M-token context, MIT licensed
New Added Aug 4, 2026
GLM-5
Why: The strongest open-weight coding model published to date: 62.1 on SWE-bench Pro against GPT-5.5's 58.6, and 81.0 on Terminal-Bench 2.1, at roughly a sixth of GPT-5.5's API price. The MIT licence carries no regional restrictions, so the weights can genuinely be self-hosted commercially, which is the reason to choose it over a closed model of similar strength.
Freemium Best Open-Weight Coder Visit
Thinking Machines' 975B Apache-2.0 model that takes text, images and audio natively
New Added Aug 4, 2026
Inkling is the first open-weights model from Thinking Machines Lab, released 15 July 2026 under Apache 2
Why: It is the first roughly trillion-parameter open-weights model that takes audio and images natively rather than through a bolted-on encoder, and Apache 2.0 means the weights can be used commercially without asking anyone. Thinking Machines is candid that this is not the strongest model available but a base worth customising, which is a more honest pitch than most open releases make.
Free Best Open-Weight Multimodal Base Visit