Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Cj Z 🎯 Profile picture
@cj_zZZz
Jan 30 • 11 tweets • 3 min read • Read on X
Interesting!
More Open Source Models are released from China.

→ Kimi k1.5 (multimodal reasoning model)
→ Qwen 2.5 Max (1M context window)
→ Mistral Small 3 (24B model)

Let's explore all these models below:
🧵 Image
1. Kimi k1.5

Official Tweet by Kimi Account
Kimi also have a web app.

UI looks really good and results are impressive (especially the vision capabilities)

2. Qwen 2.5 Max (owned by Alibaba)

Qwen cctually launched 2 models: Here's the first one.

The Best Part about Qwen Max is it's Video generation feature.

It is too good, check these.

2nd Model Alibaba released is Qwen 2.5 VL

It has great use cases:
- Computer use
- Multi modal
- Document Parsing
- Mobile Agent
- OCR

3. Mistral Small 3

- 24B parameters
- 81% MMLU score
- 150 tokens/sec (fast latency)
- Apache 2.0 license (open source)
- Pre-trained & instructed (no synthetic data – ideal for reasoning!)✅

Mistral is competing against GPT4o-mini
Kimi is competing against GPT o1
Qwen is competing against Claude 3.5 & GPT 4o

It was a great week for Open Source AI
Thanks to Deepseek.

Today we might see OpenAI launching o3-mini
Tomorrow Gemini 2.0 by Google

Exciting times ahead.
Join - The Home of AI Coders

Write Project Dev. Documentations for any AI model and use it with any AI Coding tool.

+ Enjoy Community Exclusive Discounts on
Bolt,
Windsurf,
Lovable,
and other AI tools

Limited 40% OFF on Yearly Membership. 🎉codeguide.dev

• • •

Missing some Tweet in this thread? You can try to force a refresh
　

Keep Current with Cj Z 🎯

Cj Z 🎯 Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @cj_zZZz

Cj Z 🎯 Profile picture
Jan 25
Turn any AI Model into Reasoning Model with Deepseek r1 <thinking> Architecture.

Models like GPT4o and Sonnet 3.5 are Implementation Models

But a new breakthrough with Deepseek can make them a Reasoning model.

Here's a step-by-step Explanation:
🧵
Deepseek r1 is an open source AI model competing against o1 model of OpenAI.

Which is the Top-Notch Reasoning Model.

How Reasoning Models work?
User message → Chain of Thought steps → Output

The CoT step is the main difference right now between r1 and o1
Deepseek reasoning is realtime, fast and more like human thoughts.

o1 is slow and takes 3-4 minutes on average.

But what if we use the reasoning power of r1 with model of our choice?
Read 14 tweets
Cj Z 🎯 Profile picture
Jan 23
An Open Source AI model {Deepseek r1} launched 72 hours ago.

This Chinese AI model just broke the internet.

People are running r1 model on:
- self hosted servers
- mobile phones
- raspberry pi
- mac books

Here're all the cool experiments, you check:
🧵 Image
1. Here's the Original Tweet by Deepseek Official Account.

2. Sully has picked up nice insight: "the realtime thought process of r1 model"

Read 14 tweets
Cj Z 🎯 Profile picture
Jan 20
Deepseek:

An Open Source AI Model that outperformed:
- Llama 3.2
- Gpt4o
- o1 mini

Here's everything you need to know about Chinese Deepseek Model:
🧵 Image
DeepSeek is a Chinese-developed open-source LLM series of models.

I would say this is the first real Open Source model.

It has provided details about what data and techniques it used to train the model.

MIT licence, which means you can train and build AI commercial products.
Deepseek now have 2 flagship models:
1. Deepseek-v3
2. Deepseek r1

Deepseek v3 is is competing with GPT4o and Clade Sonnet 3.5.

v3 outperformed GPT4o and Sonnet 3.5 in MMLU benchmark (to understand human language)

and It is 10x cheaper than GPT$o and Sonnet 3.5.

Deepseek API pricing:

Input: $0.27/million tokens ($0.07/million tokens with cache hits) and output: $1.10/million tokens

Claude 3.5 Sonnet API pricing:

Input costs $3.00 per million tokens, and output costs $15.00 per million tokens.

GPT4o API Pricing:

Input costs $2.50 per million tokens, and output costs $10.00 per million tokens.

Note: Sonnet 3.5 is still the best Coding model and GPT4o is great at logic and Maths.Image
Read 8 tweets
Cj Z 🎯 Profile picture
Jan 11
Cursor AI Coding Crash Course:
🧵 Image
In last 5 months, I've built 30+ projects with AI code.

And I found that Cursor is the best AI Coding IDE for complex applications like:

- Web applications
- Mobile Applications
- Downloadable Software

So If you master Cursor, you can code anything!

Let's start...
1. Ideation/Inspiration

Start with mapping out your idea. This is a very important step.

You need to visualize the screens, features, layout in your head.

You can draw on paper or go to dribbble or behance for UI design inspirations.Image
Read 16 tweets
Cj Z 🎯 Profile picture
Jan 8
In last 45 days, I wrote 12 Tutorials on AI tools.

If you missed any of them,

Here's the Master Thread for you:
🧵 Image
1. Context Boundary in Cursor

2. Tips and Tricks for Cursor & Bolt

Read 14 tweets
Cj Z 🎯 Profile picture
Jan 6
How to Create a Modern Landing Page?

→ Plan it with CodeGuide
→ Code it with Bolt .new
→ Deploy it on Netlify

Here's a Step by Step Guide:
🧵 Image
1. Take Inspirations

You need to spend some time visualizing how your landing page would look like.

For this you can go to:
- Dribbble
- Landingfolio
- onepagelove
- saaslandingpage

Once you find a style you like. Take notes of hero section, color pallete, styling etc.
2. Use CodeGuide Template

CodeGuide will help you write:
- Project Requirements Doc for your landing page
- Landing Page Copy
- Landing Page Styling Guide (Frontend Guidelines)

We'll attach these docs to Bolt and it'll save us tokens and time and landing page will look unique. Image
Read 14 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

Send Email!

:(