AI » LLM Models & Frameworks

AI » LLM Models & Frameworks

LLM: Large Language Models

Odličan info o modelima: https://poe.com/about i detaljno: Hannibal046/Awesome-LLM: Awesome-LLM: a curated list of Large Language Model

I ovde su ispisani svi modeli: Mooler0410/LLMsPracticalGuide: A curated list of practical guide resources of LLMs (LLMs Tree, Examples, Papers) sa sve rang listom.

LLM Frameworks

LangChain

Je za sada napopularniji i stoga sam napravio sam odvojeni dokument

Psychic

je valjda slično?

Universal API for unstructured data

Šta je jebeno ovo?

Psychic is a platform for integrating with your customer’s SaaS tools like Notion, Zendesk, Confluence, and Google Drive or General purpose web scraper for all other content and syncing documents from these applications to your SQL or vector database. You can think of it like Plaid for unstructured data.

Open source self-hosted: psychicapi/psychic: Unified APIs for ingesting unstructured data. Sync documents from your customers’ SaaS tools to a SQL or vector database, where they can be easily queried by AI applications Cloud version: Psychic Dashboard

Ranije se zvao Sidekick - tool to connect to your customer’s SaaS applications like Notion, Zendesk, Confluence, and Google Drive Psychic lets you scale to dozens of knowledge base, ticketing system, and CRM integrations by connecting to one universal API.

Dakle, za pravljenje API-ja iz tvojih podataka tako da možeš da pitaš sa AI podatke


ChatArena.ai is for parallel model execution, but nothing - absolutely nothing - comes without payment.

In Chatroom at OpenRouter you can also execute requests in parallel.


NVIDIA’s Llama 3.1 Nemotron: usdcode-llama3-70b-instruct | NVIDIA NIM is the latest one.


LLM AI Leaderboards

Najrelevantniji i po mojim istraživanjima ja dobijam iste rezultate na: Chatbot Arena (formerly LMSYS): Free AI Chat to Compare & Test Best AI Chatbots

Coding - Coding redosled 2025-02: 1. Gemini-2.0-Pro (nema ga jer je experimental i free) 2. DeepSeek-R1 (*** spor) 3. o1 (skup) 4. Gemini-2.0-Flash-Thinking (nema ga jer je experimental i free) 5. o3-mini (nema ga jer je closed) 6. o1 (preskup) 7. o1-mini () 8. ChatGPT-4o () 9. Gemini 2.0 Flash () 10. Qwen2.5-Max () 11. Claude 3.5 Sonnet (***)

German Language redosled 2025-02:

  1. ChatGPT-4o,
  2. Gemini-2.0-Flash-Thinking,
  3. DeepSeek-V3,
  4. o1,
  5. Claude 3.5 Sonnet

WebDev Arena je odličan Leaderboard | WebDev Arena as AI Battle to build the best website:

WebDev redosled 2025-02

  1. Claude 3.5 Sonnet (20241022)
  2. DeepSeek-R1
  3. o3-mini-high (20250131)

Copilot Arena nije mnogo relevantno jer korisnici moraju da instaliraju VSCode plugin, pa je uzorak lošiji.

Text-to-Image Redosled 2025-02:

  1. Imagen-3.0-generate-002
  2. Recraft V3
  3. Ideogram 2.0
  4. Luma Photon

Vision je prepoznavanje i analiza slika Redosled 2025-02: Gemini-2.0-Flash-Thinking, Gemini-2.0-Pro, Gemini-2.0-Flash, ChatGPT-4o

Coding

Aider LLM Leaderboards | aider Coding with Llama 3.1, new DeepSeek Coder & Mistral Large | aider RepairBench: Leaderboard of Frontier Models for Program Repair


Reasoning Models

thinking models

Moonshot Kimi k1.5 - After DeepSeek-R1, Kimi k1.5 model by Chinese startup Moonshot AI outshines OpenAI-o1 | Technology News - The Indian Express

基本信息 - Moonshot AI 开放平台

Not in API still, samo ovde: Kimi.ai - AI Assistant by Moonshot AI

date 20. May 2023 | modified 13. Feb 2025
filename: AI » LLM » Models