CryptoMaid加密女仆お嬢様 .stand|1月 29, 2026 03:38
AI industry panorama, please find your own position. It's enough to master two of each of these four layers
When we talk about AI, we are talking about:
1. Bottom level computing power/chips → NVIDIA (H100/H200/B200/Blackwell series) AMD(MI300X/MI325X/MI355X)、 Ascend (910B/910C/920), Groq LPU, Intel Gaudi 3/Habana, Google TPU v5p/v6, Cerebras Wafer Scale, SambaNova, Tenstorrent, Boren/Cambrian/Muxi/Moore Thread (mainstream in the domestic camp)
2. Foundation Models
→ Text/Universal Large Models: Claude 4 Series, Grok 3/Grok 4, GPT-4o/o1/o3 Series, Gemini 2.0/2.5, Llama 4 Series Qwen 2.5 / Qwen 3、DeepSeek V3 / R1、Mistral Large 2 / Pixtral、Yi-1.5 / Yi-2、GLM-4 / ChatGLM
→ Multimodal/Visual: Qwen-VL、Llama 4 Vision、Gemini 2.0 Flash、Claude 4 Vision、Grok Vision、Phi-4-multimodal
→ Video/Generation: Kling 2.0 underlying DiT, Runway Gen-4 underlying model, Luma Dream Machine underlying, Pika 2.0 DiT, Sora (if already open) Vidu、Haiper、Luma Ray2、Stable Video 4D
→ Audio/Music/Voice: ElevenLabs Turbo v3、Udio v2、Suno v4、Qwen-Audio、Whisper large-v3、SeamlessM4T v2
3. Intermediate engineering tools/middleware/development&inference framework
→ Inference Acceleration Engine: vLLM、SGLang、TensorRT-LLM、TGI(Text Generation Inference)、LMDeploy、Ollama( Local priority), llama. cpp/gguf ecosystem MLC-LLM、Aphrodite Engine、TabbyAPI
→ Workflow/Chain/Organization: LangChain / LangGraph、LlamaIndex、Haystack 2、Flowise、Dify、CrewAI、AutoGen、Langflow、Swarm(OpenAI)、Semantic Kernel( Microsoft) DSPy
→ Local/Self hosted: Ollama, LM Studio, GPT 4All, Anything LLM, LocalAI
→ ComfyUI (Image/Video Generation Workflow), Autotic1111 (Stable Diffusion Ecology) InvokeAI、Fooocus
→ Agent/tool calling frameworks: LangGraph, CrewAI, AutoGen, OpenAI Swarm, BabyAGI/GodMode class projects
Quantization/compression/distillation tools: bitsandbytes, AutoGPTQ, AWQ, HQQ, llama.cpp quantization Unsloth、Axolotl
4. Upper level applications/packaging tools/terminal products
→ Code/Programming Category: Cursor, Windsurf, GitHub Copilot (New Version), Codeium, Tabnine, Amazon Q Developer, Reply Agent, Devin Category Products
→ Chat/Universal Web App: http://Claude.ai 、 http://(Grok.com) / http://(x.com)、ChatGPT、 http://Gemini. (google.com)、Perplexity、Poe、http://(You.com)、Le Chat 、 Tongyi Qianwen, Doubao Kimi
Image generation: Midjourney, Flux. 1 web version/ http://Fal.ai 、 Ideogram、 http://Leonardo.Ai 、 Playground v3、Recraft、Adobe Firefly
→ Video generation webpage: Kling AI、Runway Gen-4、Pika 2.0、Luma Dream Machine、Haiper、Viggle、Gen-3 Alpha(Runway)、Sora(OpenAI If already open)
→ Voice/Dubbing/Music: ElevenLabs, HeyGen (Digital People) Synthesia、Udio、Suno、Voicify、Respeecher
→ AI short play/delivery/content factory: various Runway+Kling+Pika+ElevenLabs combination factories, HeyGen+Kling short play assembly line, Viggle+Pika dance delivery video, TikTok/Tiktok/Little Red Book AI content workshops, digital live broadcast delivery tools (mainly Alibaba, Tencent, Baidu)
Share To
Timeline
HotFlash
APP
X
Telegram
CopyLink