Models · Source-aware through June 2026
Open-weight and open-source AI models
Search and filter model families and featured models by task, hardware, workflow, source status, license posture, and practical deployment fit. OpenSourcesAI separates open-source software projects from open-weight model releases so builders can check the exact model card before production or commercial use.
Openness labeled clearly
Open-weight and open-source status is separated so you can verify the exact license before production use.
Hardware fit first
Every model surfaces VRAM and runtime needs alongside the description.
Decision-focused filters
Narrow by task, capability, or openness before comparing model cards.
Start here
Choose models by fit, not hype
The fastest path is hardware fit first, then task fit, then a small test before committing to a stack.
Hardware fit
Start with what your machine can run
Use the compatibility checker before falling in love with a model that needs more VRAM than your setup has.
Check hardware →Task fit
Choose by workflow first
Pick different models for coding, RAG, summarization, vision, audio, and reasoning instead of chasing one generic winner.
Read model guide →Test prompts
Try lightweight models in browser
Use the Playground for quick prompt checks before moving to Ollama, LM Studio, Open WebUI, or vLLM.
Open Playground →Showing 14 featured models and 15 family hubs.
Browse by model type
Filter by openness and capability
License note: model openness varies by checkpoint. Treat directory labels as a starting point and verify the official model card, license file, provider terms, redistribution rights, hosted-use limits, and derivative-use rules before commercial deployment.
Featured 2026
Featured model profiles
Reasoning
DeepSeek-V4-Pro
Frontier open-weight DeepSeek model positioned for reasoning-heavy coding, agent, and long-context work.
Best for: Teams comparing frontier-style open-weight reasoning and coding models against hosted closed models.
Verified/updated 2026
Review model →Chat
Qwen3 235B A22B
Flagship open-weight Qwen3 MoE model often chosen for serious reasoning, coding, multilingual work, and agent experiments.
Best for: Builders testing frontier-style open-weight reasoning and coding in hosted or multi-GPU environments.
Verified/updated 2026
Review model →Agents
Kimi K2.6
Current Kimi family model for agentic coding, tool-use behavior, and long-context workflow evaluation.
Best for: Builders testing agentic coding, tool calling, long-context planning, and workflow automation.
Verified/updated 2026
Review model →Code
GLM-5.2
GLM-5.2 is a 753B MoE model requiring multi-GPU or distributed inference for full deployment.
Best for: GLM-5.2 is Z.ai's large mixture-of-experts model, aimed at coding, agentic tool-use, and very long context work rather than single-GPU home setups.
Check exact model card
Review model →Reasoning
MiMo-V2.5-Pro
Xiaomi MiMo frontier model for reasoning, coding, and long-context AI application testing.
Best for: Teams comparing newer Chinese open-weight frontier models for reasoning, code, and long-context tasks.
Verified/updated 2026
Review model →Chat
Gemma 4
Google Gemma family entry for open-weight testing across efficient local, app, and multimodal workflows.
Best for: Developers evaluating Google-backed open-weight models for efficient local apps, hosted prototypes, and multimodal workflows where supported.
Verified/updated 2026
Review model →
Reasoning
MiniMax M3
MiniMax's current flagship: a 427B-parameter sparse mixture-of-experts model (MiniMaxM3SparseForConditionalGeneration) accepting both image and text input, with a 1,048,576-token (~1M) context window.
Best for: Teams evaluating frontier-scale open-weight reasoning/coding models with genuine multimodal (image) input and a ~1M-token context window, willing to work within a named commercial license rather than a standard OSI one.
Verified/updated 2026
Review model →Multimodal
Llama 4 Scout
Open-weight Llama 4 model positioned for multimodal and long-context workflows.
Best for: Teams evaluating Llama-family models for multimodal assistant, long-context, and application workflows.
Verified/updated 2026
Review model →Multimodal
Llama 4 Maverick
Open-weight Llama 4 model positioned for multimodal, reasoning, and general assistant workflows.
Best for: Builders comparing current Llama-family models for assistant, multimodal, and reasoning-oriented workflows.
Verified/updated 2026
Review model →Chat
gpt-oss-120b
Larger gpt-oss model for users evaluating high-capacity open-weight deployments.
Best for: Teams evaluating high-capacity local, self-hosted, or developer-controlled model deployments.
Verified/updated 2026
Review model →Chat
gpt-oss-20b
Smaller gpt-oss model for local and more accessible open-weight deployments.
Best for: Builders evaluating more accessible open-weight deployments for local apps, prototypes, and controlled workflows.
Verified/updated 2026
Review model →Audio
Whisper Large V3
Open speech recognition model commonly used for transcription and multilingual audio workflows.
Best for: Builders adding local transcription, podcast processing, meeting notes, or audio translation.
Verified/updated 2026
Review model →Embedding
bge-m3
Multilingual BGE embedding model that supports dense retrieval, lexical retrieval, and multi-vector retrieval in one checkpoint.
Best for: RAG teams that want one embedding model for multilingual search, long documents, and retrieval experiments beyond plain dense vectors.
Verified/updated 2026
Review model →Embedding
e5-mistral-7b-instruct
Instruction-tuned E5 embedding model built on a Mistral 7B backbone for text embedding and retrieval tasks.
Best for: Teams comparing larger instruction-tuned embedding models for retrieval quality when smaller E5 checkpoints are not enough.
Verified/updated 2026
Review model →Browse by family
Model family hubs
BAAI
BGE
BGE models cover embeddings, reranking, retrieval, semantic search, and vector database workflows for RAG builders.
Family hub · Source-aware 2026
Family hub →DeepSeek
DeepSeek
DeepSeek models cover reasoning, coding, distilled local variants, and open-weight assistant workflows.
Family hub · Source-aware 2026
Family hub →Microsoft / intfloat
E5
E5 models are widely used for multilingual embeddings, semantic search, retrieval, low-overhead indexing, and RAG pipelines.
Family hub · Source-aware 2026
Family hub →Gemma
Gemma models are useful for efficient local, app, and multimodal workflows, with small-to-mid-size variants that are practical for developers.
Family hub · Source-aware 2026
Family hub →Z.ai
GLM
GLM models from Z.ai are used for agentic engineering, tool use, coding, and reasoning workflows.
Family hub · Source-aware 2026
Family hub →OpenAI
gpt-oss
OpenAI gpt-oss models are open-weight reasoning models for local, self-hosted, and developer-controlled AI workflows.
Family hub · Source-aware 2026
Family hub →Moonshot AI
Kimi
Kimi models from Moonshot AI focus on agentic coding, tool use, long-context reasoning, and workflow automation.
Family hub · Source-aware 2026
Family hub →Meta
Llama
Meta Llama models are widely supported open-weight options for local AI stacks, multimodal workflows, assistant prototypes, and guardrail experiments.
Family hub · Source-aware 2026
Family hub →Xiaomi
MiMo
Xiaomi's MiMo family is a newer open-weight model family for reasoning, coding, long-context, and agent experiments.
Family hub · Source-aware 2026
Family hub →MiniMax
MiniMax
MiniMax models are evaluated for long-context reasoning, coding, tool use, agents, and productivity workflows.
Family hub · Source-aware 2026
Family hub →Mistral AI
Mistral
Mistral models cover multilingual workflows, coding, MoE architectures, vision-language experiments, and efficient local or hosted deployments.
Family hub · Source-aware 2026
Family hub →Meta
Muse
Muse is the model family from Meta Superintelligence Labs: Muse Glimmer is the open-weight, locally runnable member, distilled from Muse Spark, the hosted flagship.
Family hub · Source-aware 2026
Family hub →Microsoft
Phi
Microsoft Phi models focus on small language models, edge deployment, reasoning efficiency, and low-resource local experiments.
Family hub · Source-aware 2026
Family hub →Alibaba Qwen
Qwen
Qwen models are strong choices for multilingual chat, coding, math, vision-language, and local developer workflows.
Family hub · Source-aware 2026
Family hub →OpenAI
Whisper
Whisper models are used for ASR, transcription, subtitles, podcast processing, meeting notes, and multilingual audio.
Family hub · Source-aware 2026
Family hub →Next step
Turn a model choice into a working stack.
After you shortlist a model, test prompts in the Playground and use stack recipes to connect it to a chat UI, RAG workflow, coding assistant, or serving layer.
For builders
Want a model or family hub reviewed here?
Suggest a missing model family, submit an update, or sponsor a relevant placement for builders comparing model options and deployment fit. For sponsorship options, advertise with us. For submissions or corrections, use the submit page.