The AI model collection / October 2026
Explore 57 AI labs. Pick up to four cards to compare context, capabilities, and cost.
NVIDIA vision-language reasoning model for text, image, and video understanding and agentic tool use
Latest pulls
994 cards
October 17 update of DeepSeek V4.1 Flash with improved reasoning, coding, agentic tasks, and visual understanding
NVIDIA vision-language reasoning model for text, image, and video understanding and agentic tool use
Image editing model from Black Forest Labs.
Google's flagship reasoning model for complex software engineering, deep math and science, and long-horizon autonomous agents
Image editing model from Black Forest Labs, 5.5B parameters.
Muse Wonder is a frontier multimodal reasoning model from Meta for personal and professional agents, long-horizon coding, and computer use. It combines Active Memory, which carries relevant context across conversations, with Active Focus, which keeps it working through blockers toward a goal.
Language model from Liquid AI, 24B parameters.
Text-to-image model from Qwen, 7.1B parameters.
Vision-language model from OpenBMB, 35B parameters.
Fast Claude model for responsive assistance, classification, and lightweight agents
Mistral's 1T-parameter multimodal MoE for reasoning, coding, and agentic work
Vision-language model from OpenBMB, 1.6B parameters.
Vision-language model from Liquid AI, 3.1B parameters.
Cost-efficient video model for text-to-video and image-to-video generation
Near-Astra performance for complex coding, computer use, and professional work at a lower cost.
Fast Claude model for everyday coding, agents, and knowledge work
MiMo Pro model for multimodal coding agents and long-context automation
MiMo Flash model for multimodal coding agents and long-context automation
OpenAI model for complex coding and agentic workflows
OpenAI's most efficient model for focused, high-volume tasks
Claude model for long-running agentic coding and knowledge work
MiMo pro model for strong multimodal reasoning and agent execution
Language model from Xiaomi, 1T parameters.
Language model from Xiaomi, 311B parameters.
xAI's frontier model for long-running agents, coding, knowledge work, and visual projects
Speech recognition model from Edge0, 4.1B parameters.
Text classifier from Convai Innovations, 322M parameters.
Text classifier from Convai Innovations, 421M parameters.
High-speed GLM-5.3-Flash serving option for coding and agent workflows
Qwen omni model for text, vision, audio, and multimodal agent tasks
Text-to-image model from Ant Group, 6.2B parameters.
Language model from XingChen-AGI, 31B parameters.
StepFun's next-generation flagship base model for coding and professional knowledge work, with native text, image, and video input and a 1M-token context window
Vision-language model from Meta, 822B parameters.
Text-to-image model from Qwen, 7.1B parameters.
Embedding model from Google, 744M parameters.
Vision-language model from DeepSeek, 763B parameters.
Text-to-audio model from Multimodal Art Projection, 3.6B parameters.
Open-weight translation model for more than 50 languages and locale variants
Mercury 2.5 is the fastest reasoning LLM, and the latest diffusion LLM (dLLM) from Inception
Language model from OpenBMB, 324M parameters.
Language model from OpenBMB, 2.5B parameters.
Vision-language model from zidongtaichu, 9.8B parameters.
GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research, and document creation.
Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It improves long-horizon agent collaboration, instruction following, and coding efficiency relative to Muse Spark 1.2.
Muse Spark 1.3 is a multimodal reasoning model from Meta for long-running agentic, multi-agent, and coding workflows. It improves long-horizon agent collaboration, instruction following, and coding efficiency relative to Muse Spark 1.2.
Google's most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows
Claude model for demanding reasoning and long-horizon agentic work
Every spec comes from models.dev, an open, MIT-licensed database of AI models. The latest pull held 9,048 listings across 232 providers. Resellers list the same model many times over, so each model appears once here, under the lab that made it.
A scheduled job pulls the data every day, checks it, and republishes the site. New models show up as new cards without anyone editing this page. The data last changed on Oct 18, 2026.
Licenses for open-weight models come from each model’s Hugging Face card: 97 of 156 matched directly. The rest show the lab’s usual license and say so on their page.
The same daily job scans 9,133 repos across 61 official lab accounts on Hugging Face. Any model with at least 100,000 downloads in the past month becomes a card, 687 so far, and stays in the binder once it’s in. Quantized and converted copies are left out, so each model appears once.
Quality comes from two public leaderboards, both CC BY 4.0: Epoch AI’s Benchmarking Hub and LMArena’s leaderboard dataset. The same daily job matches them to cards by name, 229 so far, and each model page shows the name a board lists it under.
Modeldex isn’t affiliated with any AI lab. Model names belong to their makers, and every card links to the lab’s own docs.