AI Models

Compare the leading AI models across Global, China and Japan — specs, heat scores, API console access and official links, all in one place.

Showing 52–79 of 79 modelsTotal 79 models
MiniMax M3
60
Native multimodal, 1M context, MSA cuts cost 20x, built for multi-step complex tasks.
Qwen3.8 27B
60
Alibaba’s Qwen team finally released the model the open-source community had been waiting for: Qwen3.8-27B.What caught my attention was the size: 27B parameters. That is still small enough to be realistic for local deployment on high-end consumer hardware, especially when quantized.
Tencent: Hy3
59
Tencent Hunyuan’s advanced MoE language model optimized for reasoning, coding, long-context understanding, and AI Agent applications.
MiMo-V2.5
52
Xiaomi MiMo’s next-generation multimodal AI model, combining text, image, video, and audio understanding with advanced reasoning and Agent capabilities.
Ring-2.6-1T
51
Ring-2.6-1T is a 1T-parameter-scale thinking model with 63B active parameters...
Nemotron 3.5 Content Safety
42
NVIDIA’s open safety model that automatically detects AI content risks and helps protect large language model applications with safer, more compliant AI experiences.
Granite 4.1 8B
41
IBM’s next-generation 8B parameter open-source model, optimized for tool calling, instruction following, coding assistance, and enterprise AI applications.
Cohere North Mini Code
41
Cohere North Mini Code is Cohere's first agentic coding model and the debut of its North family...
Ling-2.6-flash
39
InclusionAI’s efficient open-source language model optimized for fast responses, AI Agent workflows, reasoning tasks, and high token efficiency.
Fugu Ultra
38
Sakana AI’s high-performance multi-agent AI system that coordinates multiple specialized Agents to solve complex tasks with improved accuracy and reliability.
KAT-Coder-Pro V2.5
37
A next-generation AI coding model designed for software engineering tasks, supporting code development, project maintenance, and intelligent Agent workflows.
Laguna S 2.1
36
Poolside’s next-generation MoE AI model designed for agentic coding, long-context reasoning, and complex software engineering tasks.
Nemotron 3 Nano Omni
36
NVIDIA’s open multimodal AI model designed for text, image, video, and audio understanding, optimized for AI Agents, reasoning, and enterprise applications.
KAT-Coder-Air V2.5
35
A lightweight AI coding model optimized for software engineering, supporting code generation, project analysis, debugging, and Agent-powered development workflows.
Step 3.7 Flash
34
The high-performance Flash model launched by StepFun for AI agents and developers features native visual understanding, multimodal reasoning, tool calling, and code generation capabilities...
Sakana Namazu
33
What caught my attention about Namazu wasn’t the “Japanese LLM” label. It was what Sakana AI actually changed. The model is built on Kimi K2.6. Sakana didn’t try to build a new foundation model from scratch. Instead, it focused on Japanese post-training, tool use, and packaging the whole thing into an API that developers can actually plug into a product.
Aion-3.0
32
Laguna M.1
31
Poolside’s next-generation MoE language model designed for AI coding agents, software engineering, long-horizon tasks, and advanced developer workflows.
Laguna XS 2.1
28
Poolside’s lightweight MoE coding model designed for local AI Agents, software development, and efficient long-horizon programming workflows.
Aion-3.0-Mini
25
A lightweight AI model by AionLabs focused on roleplay, storytelling, creative writing, and high-quality text generation with efficient performance.
Nex-N2-Pro
21
An intelligent Agent model built for real-world tasks, with enhanced reasoning, coding, and autonomous execution capabilities.
Nex-N2-Mini
20
Nex AGI’s open-source Agent model designed for coding, tool use, deep research, and advanced productivity workflows.
Perceptron Mk1
16
A multimodal AI model built for image, video understanding, and physical AI tasks with advanced visual reasoning and structured outputs.
Thinking Machines: Inkling
16
An open-weight multimodal AI model by Thinking Machines Lab, designed for AI Agents, coding, and advanced intelligent applications.
Thinking Machines: Inkling Small
13
A lightweight open-weight AI model from Thinking Machines Lab, designed for AI Agents, coding, multimodal understanding, research, and customizable enterprise AI applications.
Muse Spark 1.2 Contributor
The first thing that caught my attention about Muse Spark 1.2 Contributor wasn’t the low price. It was the question behind it: why is Meta willing to make it this cheap? The answer is in the terms. If you let Meta use your interactions to improve its models, you get much cheaper API access.
DeepSeek V4 Flash Vision Exp
DeepSeek launched a new model on its API platform on August 21: DeepSeek-V4-Flash-Vision-Exp. The name is a mouthful, but the idea is simple: V4-Flash can now see images.
GLM 5.3 Flash
Ox Alpha topped OpenRouter on day one and broke the token record.Zhipu confirmed it as GLM-5.3-Flash on Aug 26.

Frequently Asked Questions

How many AI models are listed?

Aivoax continuously collects and updates information about different AI models.
The platform covers major international models such as GPT, Claude, Gemini and DeepSeek, as well as Chinese-developed models including Qwen, Ernie and Hunyuan.
Models are organized by category, including general-purpose language models, multimodal models and code-focused models, making it easier to find options based on different needs.
Since the AI industry changes quickly, we regularly review new releases and update information about existing models.

How are models ranked?

Aivoax organizes model information from different perspectives, including regions and usage scenarios.
Users can explore models based on global, Japan and China views, as well as different needs such as enterprise use, learning and developer workflows.
Rankings are provided as a reference for understanding current attention and popularity. They do not represent official performance rankings.
When choosing a model, factors such as use case, pricing, speed and available features should also be considered.

Where can I check a model's specs and API access?

Each model page includes key information such as model overview, context length, pricing details, open-source status, release information and official links.
For developers, we also provide API documentation links to help with evaluation and integration.
Because AI models are updated frequently, specifications may change over time. Please check official documentation before production use.

What is the heat score?

The Heat Score is a reference indicator used by Aivoax to show how much attention an AI model is receiving.
It is based on publicly available signals such as search trends, community discussions and user interest.
A higher score indicates greater current attention, but it does not necessarily mean the model has better performance.
Different models are suitable for different tasks, including coding, content creation and business applications.

What's the difference between a model and a tool?

An AI model is the underlying technology that provides AI capabilities, such as GPT, Claude and Qwen. Developers usually access these models through APIs.
An AI tool is a finished application built using one or more AI models, such as chat assistants, writing tools and image generation services.
In simple terms, models provide the core capability, while tools turn that capability into products that users can directly use. Aivoax organizes both separately for easier discovery.

What else is on a model's detail page?

Model detail pages include more than basic specifications. They also provide information about suitable use cases, key features, limitations and related models.
These details are designed to help users compare different options more efficiently.
There is no single AI model that works best for everyone. The right choice depends on your goals, budget and usage environment.