Back to directory
GPT-6OpenAI's frontier model family powering ChatGPT.
Nano BananaGoogle's image generation and editing model.
DALL·E 3OpenAI's retired image model, succeeded by the GPT Image family.
AutoGPTOpen-source AI agent platform, reborn from the 2023 autonomous-agent phenomenon.
Gen-2Runway's retired early video model; the current line is Gen-4.5.
PaLM 2Google's legacy LLM, superseded by the Gemini family.
CohereLarge language model platform for building AI products.
Seedream 5.0 ProByteDance Seed's multimodal image model: native interactive editing and Chinese text rendering, from CNY 0.30 per image on Volcengine Ark.
Qwen-Image 2.1Alibaba Qwen's open-weight image model: 7B unified generation and editing, native RGBA transparency, 10 references - research license, non-commercial.
Qwen3.5Alibaba's strongest open base model: 397B MoE / 17B active, natively multimodal, API from CNY 0.8.
MiniMax M3Multimodal coding flagship: 1M context, $0.30/$1.20 permanent half price, MIT open weights.
Hunyuan Hy3Tencent's open flagship: 295B MoE / 21B active at CNY 1/4 - extreme value, Apache 2.0.
Ideogram 4.5Built for drift-free multi-turn precise editing: native 2K, four quality modes, Precise Edit from $0.008 per image.
MAI-Image-2.5 ProMicrosoft's highest-fidelity in-house image model: public preview on Foundry at $5 per 1M text input and $106 per 1M image output tokens.
GLM-5Zhipu Z.ai's open flagship: 744B MoE under MIT, top open-source tier for reasoning and coding.
Claude ModelsAnthropic's model lineup: Fable 5 flagship, Opus 4.8 agentic coding, Sonnet 5 balanced, Haiku 4.5 fast.
FLUX 3 ImageBlack Forest Labs' image flagship: native 4K, 10 reference images, bounding-box control; Klein runs on 8GB VRAM.
Stable Diffusion 4SD4 in two tiers: Ultra with native 4096-squared output and a dedicated text module, open weights.
GPT Image 2.5OpenAI's current image model: flare and sunburst tiers at $8/$30 per million image tokens.
Nano Banana ProGoogle's flagship image model (Gemini 3 Pro Image): up to 4K, 14 reference images, multilingual text rendering, API at $0.134 per 1K/2K image.
Recraft V4A designer-first image model: the only one producing editable production-grade SVG, raster from $0.035 and vector from $0.08, Pro at native 2048 squared.
Krea 2Krea's first in-house foundation image model: 12B DiT open weights, API from $0.03 (Medium) and $0.06 (Large).
Grok Imagine 2.0xAI image generation and editing model: 1K/2K output, edits with up to 5 reference images, from $0.04 per image at 1K.
Qwen-Image 3.0 ProQwen's image flagship for complex layouts and accurate text: $0.04 at 1K and $0.075 at 2K internationally; CNY 0.25 and 0.5 domestically.
Doubao ModelsByteDance's family of language, vision, speech, and video generation models.
HuggingFaceAI model development community.
Llama 4Meta's open-weight family: Scout with 10M context, Maverick with 1M.
OllamaRun Llama and other large language models locally.
ERNIEBaidu's knowledge-enhanced large language model family.
Tencent HunyuanTencent's large language model family for Chinese content, reasoning, and task execution.
iFlytek Spark MaaSAI platform by iFlytek for trying, calling, deploying, and fine-tuning large models.
Kimi K3Moonshot's flagship: 1M context, $3/$15 per million tokens, open weights for self-hosting.
SenseChatLarge language model and AI chat assistant by SenseTime.
InternLM Shusheng ModelSeries of open AI models from Shanghai AI Laboratory (InternLM).
ModelScope CommunityAI model community by Alibaba DAMO Academy with 300+ open-source models.
Aivinla Wujie Weilai AI ModelAI applications and elastic computing platform.
MiracleVision AI ModelAI vision large model by Meitu supporting AI image, design, and video creation.
DataifyData acquisition API, high-quality datasets, integrated provisioning of proxy resource services.
Grok 4.7xAI's flagship: 500K context, $2/$6 per million tokens, real-time insight from X.
GemmaOpen lightweight AI model family by Google.
LobeSimple and free machine learning model training tool.
Scale AIAI machine learning annotation training platform.
DeepSeek V4The value king: 1M context, Flash at $0.14/$0.28, MIT-licensed open weights.
Cherry StudioOpen source universal AI client assistant.
AnythingLLMOpen source full-stack AI client supporting on-premises deployment and API integration.
Chatbox AIOpen source AI client assistant supporting multiple mainstream AI models.
JanA tool to run large models locally and have AI conversations, free and open source.
AgentGPTOpen source project to assemble, configure and deploy autonomous AI in a browser.
GradioOpen source Python library for building machine learning model UI interfaces.
DeepSpeedMicrosoft open source low-cost implementation of ChatGPT-like model training.
Watsonx.aiIbm's enterprise generative artificial intelligence and machine learning platform.
ReplicateRun open source machine learning models online.
Evidently AIOpen source machine learning model monitoring and testing tool.
Mistral Large 3Mistral's open-weight flagship: 675B MoE, 256K context, Apache 2.0.
Amazon Nova 2AWS's in-house model family: Lite for reasoning, Pro for complex work, Sonic for voice.
Gemini 3Google's third-generation family: Pro at $2/$12 - the cheapest frontier tier, 1M context.
AI-NAV DIRECTORY
AI Models
56 tools available
This AI model directory covers open-source large language models, foundation models, vision and speech models, plus model download and demo platforms. Compare parameter size, context length, Chinese performance, inference requirements, licenses, quantized releases, and local deployment needs. Developers and researchers should review official model cards, hardware memory requirements, safety limitations, and commercial terms before downloading or deploying a model rather than relying on benchmark rankings alone.
