Find any AI model

Every model worth trying from the past year — and everything in Ollama and LM Studio — explained simply. Download it, or open it in Lockedmind. Nothing you type ever leaves your device.

Checking this browser…

Your memory

Not sure how much memory you have? On a Mac: Apple menu › About This Mac. On Windows: Settings › System › About.

Opens in Lockedmind Fits my computer Original makers only Get it Released Sort

Start here

Popular models

Searching and filtering need JavaScript; here are the most popular models.

Llama 3.1 8B (Meta · 8.03B · chat) — Meta's well-known 8B model: reliable chat and writing in many languages.
Gemma 4 31B (Google · 31.27B · chat, photos) — Google's most capable open model. Needs a powerful computer.
Qwen3 0.6B (Alibaba (Qwen) · 0.75B · chat) — Tiny, yet it can think step by step before it answers. Good for trying things out on an older computer.
Llama 3.2 1B (Meta · 1.24B · chat) — A fast, light everyday assistant that works on most laptops. The best place to start on an older computer.
Qwen3 8B (Alibaba (Qwen) · 8.19B · chat) — Smart and versatile, with optional step-by-step thinking. Needs a strong computer to run in the browser.
Llama 3.2 3B (Meta · 3.21B · chat) — A solid everyday assistant: clearly smarter than the 1B, and still quick.
gpt-oss 20B (OpenAI · 20.91B · chat) — OpenAI's open model: strong reasoning. It only just fits a 16 GB computer.
gpt-oss 120B (OpenAI · 116.83B · chat) — OpenAI's biggest open model. Needs a workstation with 80 GB of memory or more.
Qwen3.5 9B (Alibaba (Qwen) · 9.65B · chat, photos) — The largest model that runs in a browser here, and the smartest — if your computer can handle it.
Gemma 4 26B (Google · 25.81B · chat, photos) — A 'mixture of experts': big-model smarts that runs faster than its size suggests. Sees images.
Gemma 4 E4B (Google · 8B · chat, photos, omni) — What Lockedmind for Mac runs: quick and clever, and it can read photos and documents.
Gemma 3 1B (Google · 1B · chat) — Google's smallest model: friendly writing and good with many languages, in well under 1 GB.
Gemma 3 4B (Google · 4.3B · chat, photos) — Small, but it can read images. A great first model for photos and screenshots.
Gemma 4 12B (Google · 11.96B · chat, photos, omni) — More knowledge and sharper reasoning than E4B, and it sees images too.
Qwen3.5 4B (Alibaba (Qwen) · 4.66B · chat, photos) — The newest mid-size Qwen, with sharper answers than Qwen3 4B at a similar size.
Qwen2.5 7B (Alibaba (Qwen) · 7.62B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Gemma 4 E2B (Google · 5.12B · chat, photos, omni) — Google's newest small model. It understands images and audio as well as text.
Qwen3.8 27B (Alibaba (Qwen) · 27.78B · chat, photos) — Qwen3.8 delivers substantial gains across coding, professional work, research, and long-horizon agentic tasks.
Qwen3 30B (Alibaba (Qwen) · 30.53B · chat) — Also a mixture of experts: big-model answers at small-model speed, if you have 32 GB of memory.
Bge m3 (BAAI · embed) — BGE-M3 is a new model from BAAI distinguished for its versatility in Multi-Functionality, Multi-Linguality, and Multi-Granularity.
Gemma 2 2B (Google · 2.61B · chat) — Google's popular small model, known for natural, pleasant writing.
Gemma 2 9B (Google · 9.24B · chat) — Google's larger Gemma 2, with thoughtful, well-written answers.
Qwen3 1.7B (Alibaba (Qwen) · 2.03B · chat) — A capable small all-rounder that can switch on step-by-step thinking for tricky questions.
Qwen2.5 VL 7B (Alibaba (Qwen) · 8.29B · chat, photos) — Reads documents, charts and photos in fine detail.
Qwen3.5 27B (Alibaba (Qwen) · 27.78B · chat, photos) — Large and very capable, with deep knowledge and careful reasoning.
Qwen2.5 Coder 7B (Alibaba (Qwen) · 7.62B · chat, code) — A serious coding assistant: it writes, explains and reviews real code.
Qwen3.5 0.8B (Alibaba (Qwen) · 0.87B · chat, photos) — The newest tiny Qwen: quick answers in many languages, on almost any device.
Qwen3 14B (Alibaba (Qwen) · 14.77B · chat) — A big step up in knowledge and reasoning, with step-by-step thinking.
Llama 3.3 70B (Meta · 70.55B · chat) — Meta's large model, close to the best there is. Needs 64 GB of memory.
Phi-4 14B (Microsoft · 14.66B · chat) — Microsoft's larger Phi: excellent at maths, science and careful reasoning.
Qwen3.5 2B (Alibaba (Qwen) · 2.27B · chat, photos) — The newest small Qwen, with a good balance of speed and smarts for everyday use.
Qwen2.5 1.5B (Alibaba (Qwen) · 1.54B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Qwen3 4B Instruct 2507 (Alibaba (Qwen) · 4.02B · chat) — Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
Qwen2.5 0.5B (Alibaba (Qwen) · 0.49B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Qwen3 Coder 30B (Alibaba (Qwen) · 30.53B · chat, code) — One of the best open coding models, and fast for its size.
Qwen2.5 Coder 32B (Alibaba (Qwen) · 32.76B · chat, code) — The latest series of Code-Specific Qwen models, with significant improvements in code generation, code reasoning, and code fixing.
Gemma 3 27B (Google · 27.43B · chat, photos) — The current, most capable model that runs on a single GPU.
Qwen3 32B (Alibaba (Qwen) · 32.76B · chat) — Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
Qwen3.6 35B A3B (Alibaba (Qwen) · 35.95B · chat, photos) — Qwen3.6 delivers substantial upgrades in agentic coding and thinking preservation than previous Qwen models.
Qwen3.5 35B A3B (Alibaba (Qwen) · 35.95B · chat, photos) — Qwen 3.5 is a family of open-source multimodal models that delivers exceptional utility and performance.
Qwen2.5 3B (Alibaba (Qwen) · 3.09B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Qwen3 VL 8B (Alibaba (Qwen) · 8.77B · chat, photos) — The most powerful vision-language model in the Qwen model family to date.
Llama 2 7B Chat (Meta · 6.74B · chat) — Llama 2 is a collection of foundation language models ranging from 7B to 70B parameters.
Qwen3.6 27B (Alibaba (Qwen) · 27.78B · chat, photos) — Qwen3.6 delivers substantial upgrades in agentic coding and thinking preservation than previous Qwen models.
DeepSeek V3 (DeepSeek · 684.53B · chat) — A strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.
Llama 3.1 70B (Meta · 70.55B · chat) — Llama 3.1 is a new state-of-the-art model from Meta available in 8B, 70B and 405B parameter sizes.
Qwen2.5 Coder 14B (Alibaba (Qwen) · 14.77B · chat, code) — A strong coding model for real projects, if you have the memory.
Qwen3 235B A22B (Alibaba (Qwen) · 235.09B · chat) — Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models.
Qwen3 Embedding 0.6B (Alibaba (Qwen) · 0.6B · embed) — Building upon the foundational models of the Qwen3 series, Qwen3 Embedding provides a comprehensive range of text embeddings models in various sizes
Gemma 3 12B (Google · 12.19B · chat, photos) — The current, most capable model that runs on a single GPU.
Qwen2.5 72B (Alibaba (Qwen) · 72.71B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Llama 3.2 Vision 11B (Meta · 10.67B · chat, photos) — Meta's model that understands images: describe, compare and read what's in a picture.
Qwen2.5 14B (Alibaba (Qwen) · 14.77B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Qwen2.5 32B (Alibaba (Qwen) · 32.76B · chat) — Qwen2.5 models are pretrained on Alibaba's latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support.
Embeddinggemma 300M (Google · 0.3B · embed) — EmbeddingGemma is a 300M parameter embedding model from Google.
Qwen2.5 Coder 1.5B (Alibaba (Qwen) · 1.54B · chat, code) — A small coding helper: explains code, writes short functions and spots bugs.
Qwen2.5 Coder 3B (Alibaba (Qwen) · 3.09B · chat, code) — A stronger small coding helper. Its licence allows personal and research use only.
GLM 4.7 Flash (Z.ai (GLM) · 31.22B · chat) — As the strongest model in the 30B class, GLM-4.7-Flash offers a new option for lightweight deployment that balances performance and efficiency.
Qwen3.5 122B A10B (Alibaba (Qwen) · 125.09B · chat, photos) — Qwen 3.5 is a family of open-source multimodal models that delivers exceptional utility and performance.
Qwen3 Embedding 8B (Alibaba (Qwen) · 7.57B · embed) — Building upon the foundational models of the Qwen3 series, Qwen3 Embedding provides a comprehensive range of text embeddings models in various sizes