Overview
The easy-learn-ai project aims to compile a comprehensive reference of AI model information. Until July 12, 2026, all model data — text, image, and video generation models from every vendor — was crammed into three files (model.json, img.json, video.json) totaling nearly 6,000 lines of JSON, making lookup and maintenance painful.
On that date, developer lishiqi.conard restructured the catalog into 19 vendor-based files, reorganizing the entire collection by manufacturer.
Largest Vendor Files
| File | Lines | Representative model families |
|------|-------|------------------------------|
| openai.json | 981 | GPT series, o series |
| alibaba.json | 752 | Qwen series |
| zhipu-ai.json | 570 | ChatGLM series |
| bytedance.json | 518 | Seed series |
| deepseek.json | 487 | DeepSeek-R1/V3 series |
| google.json | 421 | Gemini series |
| anthropic.json | 373 | Claude series |
| moonshot.json | 365 | Kimi series |
| baidu.json | 362 | ERNIE series |
| tencent.json | 353 | Hunyuan series |
Key Observations
From capability-based to vendor-based classification
The old structure asked "what can these models do?" (text/image/video); the new one asks "where do they come from?" The post argues that in 2026, model identity is increasingly about ecosystem: choosing a model means choosing OpenAI's APIs, self-hosting DeepSeek, or Alibaba Cloud.DeepSeek: the open-source disruptor
deepseek.json lists ~20 models, from flagship DeepSeek-R1 to distillations (Distill-Qwen-1.5B/7B/14B/32B, Distill-Llama-8B/70B), each with a parent field pointing to R1 — a literal family tree. All are open source, with context windows of 64K–128K.Anthropic: the safety-first flagship
The Claude lineup (Opus 4.6, Opus 4.8, Sonnet 4.6, Haiku 4.5) emphasizes agentic coding, long-context stability, tool-trigger reliability, and honesty. Claude Opus 4.8 supports a 1M-token context window.Alibaba: hedging open and closed
Qwen3.5-Plus is open source (397B total params, 17B active, MMLU-Pro 87.8%, priced at 1/18 of Gemini 3 Pro), while Qwen3.7-Max and Qwen3.6-Plus are closed-source, exclusive to Alibaba Cloud Model Studio with 1M-token context.Google: blurring flagship and speed
Gemini 2.0 Flash targets high throughput and low cost with 1M context and native multimodality; Gemini 3.1 Pro scores 77.1% on ARC-AGI-2; Gemini 3.5 Flash reportedly beats 3.1 Pro on several coding and agentic benchmarks despite being a "Flash" tier.The context-window arms race
- GPT-4 (2023): 8K
- GPT-4o (2024): 128K
- Gemini 2.0 Flash (2025): 1M
- Claude Opus 4.8 (2026): 1M
- Qwen3.7-Max (2026): 1M
- Open: DeepSeek (nearly all), Alibaba (Qwen3.5), Meta (Llama), Zhipu AI (partial ChatGLM), Stability AI (image models)
- Closed: OpenAI, Anthropic, Google, Baidu, Moonshot, ByteDance (Seed)
The post argues that beyond ~1M tokens, AI shifts from a chat tool to a research assistant.
Open source vs. closed source
Model tags reveal an evolution
Tag frequency across models shows a trajectory: text generation first, then visual understanding, then tool calling as standard on mid/high-end models — with deep thinking the hottest tag of 2025–2026.Conclusion
The reorganization — from three files to nineteen, from capability to vendor — mirrors an industry shift: models are no longer standalone tools but entry points into ecosystems, flags on an expanding map of AI empires.
---
*Based on easy-learn-ai commit e6c189a.*