Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Introducing RAT: Retrieval Argument Thinking. Extract just the reasoning process from deepSeek-r1 and send it to any LLM via . Boost the second LLM's performance and gain access to missing capabilities like function calling and JSON mode. 👇

Pietro Schirano

102,147 subscribers

532,880 views • 1 year ago •via X (Twitter)

Education Science & Technology

Anya Rossi• Live Now

Private livecam show

11 Comments

Pietro Schirano1 year ago

🐀🐀🐀

Pietro Schirano1 year ago

You can install RAT as a command-line tool, allowing you to run it from anywhere by simply typing `rat`. There is also a specialized implementation designed for Claude models that leverages Anthropic's message prefilling capabilities. This version makes Claude believe the reasoning process is its own internal thought process, leading to more coherent and contextually aware responses.

CodeRabbit1 year ago

AI-first pull request reviewer with context-aware feedback, line-by-line code suggestions, and real-time chat.

Pietro Schirano1 year ago

I meant Augmented lol 🤦‍♂️

Zain1 year ago

@OpenRouterAI Shouldn't it be TAG - thinking augmented generation?

The Canaanite1 year ago

@OpenRouterAI @lateinteraction I cannot not think of DSPy when I see this, lol. It feels like pure gold.

Carlos1 year ago

@OpenRouterAI We’re so early that we can just make things up

Ryszard Rzepa1 year ago

@OpenRouterAI This is just so good. Get the smart ai do the thinking and “less smart” do the job.

YamangilAnil1 year ago

@OpenRouterAI Great! Have been waiting to see this since franken-sonnet introduction of yours lol. It is a freaking good idea. And keeping even only "original" thought token is a really powerful boost. Will try to introduce more proper CoT synthesis + instruction set to Claude as prefill..

🐧 lalo adrian morales 𝕏1 year ago

@OpenRouterAI nice!!! i like the name !

Fábio Zé Domingues1 year ago

@OpenRouterAI Ufff. That last paragraph said it all. Tks

Related Videos

By the way, you can extract JUST the reasoning from deepseek-reasoner, which means you can send that thinking process to any model you want before they answer you. Like here where I turn gpt-3.5 turbo into an absolute genius!

By the way, you can extract JUST the reasoning from deepseek-reasoner, which means you can send that thinking process to any model you want before they answer you. Like here where I turn gpt-3.5 turbo into an absolute genius!

Pietro Schirano

1,324,703 views • 1 year ago

DeepSeek R1 671B just broke speed records at 198 t/s it is now the fastest reasoning model available you can try it in coding mode on anychat soon prompt: make two fancy D3.js animated looping speedometers, first one is title deepseek R1 671B and second one is OpenAI o3-mini, deep seek should go up to 198 tokens/second and openai o3-mini should go up to 174 tokens/second

DeepSeek R1 671B just broke speed records at 198 t/s it is now the fastest reasoning model available you can try it in coding mode on anychat soon prompt: make two fancy D3.js animated looping speedometers, first one is title deepseek R1 671B and second one is OpenAI o3-mini, deep seek should go up to 198 tokens/second and openai o3-mini should go up to 174 tokens/second

AK

135,360 views • 1 year ago

The most amazing aspect of DeepSeek Open Source is that just the Reasoning Engine can be isolated and used with any other LLM. In fact you can have a mixture of Reasoning Engines cascade around a problem and then use an Operator type agent AI to function on the the results.

The most amazing aspect of DeepSeek Open Source is that just the Reasoning Engine can be isolated and used with any other LLM. In fact you can have a mixture of Reasoning Engines cascade around a problem and then use an Operator type agent AI to function on the the results.

Brian Roemmele

398,746 views • 1 year ago

Introducing Aura Checker! 🔮 - give your and someone’s X username - add any other details - X posts search powered by Exa - get analysis powered by DeepSeek R1 reasoning model provided by Groq Inc and DeepSeek - have fun :) link to demo and repo in replies 🔗 demo 👇

Introducing Aura Checker! 🔮 - give your and someone’s X username - add any other details - X posts search powered by Exa - get analysis powered by DeepSeek R1 reasoning model provided by Groq Inc and DeepSeek - have fun :) link to demo and repo in replies 🔗 demo 👇

Zaid

23,011 views • 1 year ago

🚨 DeepSeek just dropped DeepSeek R1-0528 and it's INSANE for coding Built an entire game with Nebius Studio in three turn. The improvement from previous versions is wild. - API access available - Safe to use If this is R1... DeepSeek R2 is going to be absolutely nuts. How to access 👇 1/3

🚨 DeepSeek just dropped DeepSeek R1-0528 and it's INSANE for coding Built an entire game with Nebius Studio in three turn. The improvement from previous versions is wild. - API access available - Safe to use If this is R1... DeepSeek R2 is going to be absolutely nuts. How to access 👇 1/3

AshutoshShrivastava

61,058 views • 1 year ago

INTRODUCING Nebula ☄️ The most powerful AI to read and write to every EVM chain — with advanced blockchain reasoning and execution capabilities. ◆ Query real-time onchain data ◆ Perform any onchain action ◆ Build autonomous web3 agents Get Early Access👇

INTRODUCING Nebula ☄️ The most powerful AI to read and write to every EVM chain — with advanced blockchain reasoning and execution capabilities. ◆ Query real-time onchain data ◆ Perform any onchain action ◆ Build autonomous web3 agents Get Early Access👇

thirdweb

955,999 views • 1 year ago

Introducing SYNTHETIC-1: Collaboratively generating the largest synthetic dataset of verified reasoning traces for math, coding and science using DeepSeek-R1. Join us to contribute compute towards state-of-the-art open reasoning models.

Introducing SYNTHETIC-1: Collaboratively generating the largest synthetic dataset of verified reasoning traces for math, coding and science using DeepSeek-R1. Join us to contribute compute towards state-of-the-art open reasoning models.

Prime Intellect

198,473 views • 1 year ago

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: 🧵

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: 🧵

CJ Zafir

192,262 views • 1 year ago

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

The DeepSeek-R1 paper is a gem! Highly encourage everyone to read it. It's clear that LLM reasoning capabilities can be learned in different ways. RL, if applied correctly and at scale, can lead to some really powerful and interesting scaling and emergent properties. There is more to RL than meets the eye! Here is my breakdown of the paper along with a few tests: The multi-state training might not make sense initially but they provide clues on optimizations that we can continue to tap into. Data quality is still very important for enhancing the usability of the LLM. Unlike other reasoning LLMs, DeepSeek-R1's training recipe and weights are open so we can build on top of it. This opens up exciting research opportunities. About the attached clip: the previous preview model wasn't able to solve this task. DeepSeek-R1 can solve this and many other tasks that o1 can solve. It's a very good model for coding and math.

elvis

140,600 views • 1 year ago

OpenAI people alerted that you will send personal data to DeepSeek, but you can actually rent a GPU from Hyperbolic and host your own R1 model using ollama to avoid sending data to any company. DeepSeek R1 is the true user-owned AGI. The beauty of open source!

OpenAI people alerted that you will send personal data to DeepSeek, but you can actually rent a GPU from Hyperbolic and host your own R1 model using ollama to avoid sending data to any company. DeepSeek R1 is the true user-owned AGI. The beauty of open source!

Jasper

880,229 views • 1 year ago

Google just dropped Gemini 2.5 Pro with native thinking capabilities. It outperforms OpenAI o3-mini, GPT 4.5, Claude 3.7 Sonnet, DeepSeek R1 and features 1 million context with tool use and structured outputs. 100% free on Google AI Studio and via API.

Google just dropped Gemini 2.5 Pro with native thinking capabilities. It outperforms OpenAI o3-mini, GPT 4.5, Claude 3.7 Sonnet, DeepSeek R1 and features 1 million context with tool use and structured outputs. 100% free on Google AI Studio and via API.

Shubham Saboo

26,724 views • 1 year ago

Announcing Payments MCP, the easiest way for AI agents to get onchain via x402. 🚀 It lets LLM models like Claude, Gemini, and ChatGPT gain access to onchain tools like wallets, onramp, and payments with no API key required. 🧵

Announcing Payments MCP, the easiest way for AI agents to get onchain via x402. 🚀 It lets LLM models like Claude, Gemini, and ChatGPT gain access to onchain tools like wallets, onramp, and payments with no API key required. 🧵

Coinbase Developer Platform🛡️

417,658 views • 7 months ago

The new version of DeepSeek-R1, it's an INSANE model, my god. DeepSeek Engineer v2 with function calling support drops tomorrow. I have to.

The new version of DeepSeek-R1, it's an INSANE model, my god. DeepSeek Engineer v2 with function calling support drops tomorrow. I have to.

Pietro Schirano

138,907 views • 1 year ago

Function calling is a powerful way to extend the capabilities of LLMs and AI agents by letting them use external tools. Our new short course Function calling and Data Extraction with LLMs, created with @NexusflowX and taught by Jiantao Jiao and Venkat, demonstrates how to prompt LLMs to form calls to external functions. You'll work with NexusRavenV2-13B, a 13B parameter open-source model that excels in function calling tasks while still being small enough to host locally. Learn to use function calling to extract structured data from unstructured text and access web APIs, and build an end-to-end application that processes customer service transcripts. You'll learn how to build LLM-powered applications that can analyze feedback, automate data entry, and enhance search. Please get started here:

Function calling is a powerful way to extend the capabilities of LLMs and AI agents by letting them use external tools. Our new short course Function calling and Data Extraction with LLMs, created with @NexusflowX and taught by Jiantao Jiao and Venkat, demonstrates how to prompt LLMs to form calls to external functions. You'll work with NexusRavenV2-13B, a 13B parameter open-source model that excels in function calling tasks while still being small enough to host locally. Learn to use function calling to extract structured data from unstructured text and access web APIs, and build an end-to-end application that processes customer service transcripts. You'll learn how to build LLM-powered applications that can analyze feedback, automate data entry, and enhance search. Please get started here:

Andrew Ng

110,420 views • 2 years ago

Data preprocessing is critical for building effective RAG systems. Our new short course, Preprocessing Unstructured Data for LLM Applications, taught by Matt Robinson of Unstructured, demonstrates important but sometimes overlooked aspects of RAG systems: - How to extract and normalize content from diverse formats like PDF, Powerpoint, and HTML to expand your LLM's knowledge - Enriching data with metadata to enable more powerful retrieval and reasoning - Applying document layout analysis and vision transforms to process embedded images and tables Then you’ll use all these skills and build a RAG bot that draws from a corpus that includes PDF, PowerPoint, and Markdown documents. Please sign up here:

Data preprocessing is critical for building effective RAG systems. Our new short course, Preprocessing Unstructured Data for LLM Applications, taught by Matt Robinson of Unstructured, demonstrates important but sometimes overlooked aspects of RAG systems: - How to extract and normalize content from diverse formats like PDF, Powerpoint, and HTML to expand your LLM's knowledge - Enriching data with metadata to enable more powerful retrieval and reasoning - Applying document layout analysis and vision transforms to process embedded images and tables Then you’ll use all these skills and build a RAG bot that draws from a corpus that includes PDF, PowerPoint, and Markdown documents. Please sign up here:

Andrew Ng

150,317 views • 2 years ago