Benjamin Klieger
@BenjaminKlieger
Followers
2,845
Following
318
Media
152
Statuses
817
@stanford ‘28, building apps @groqinc
Newest projects on Github
Joined January 2023
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Milton
• 998115 Tweets
Florida
• 902770 Tweets
DeSantis
• 271319 Tweets
60 Minutes
• 249736 Tweets
#ファミマのスイーツパン誕生
• 123763 Tweets
Chiefs
• 111678 Tweets
Saints
• 84137 Tweets
サンデー
• 62084 Tweets
Yankees
• 40794 Tweets
#HaryanaElectionResult
• 38472 Tweets
Royals
• 38041 Tweets
Carr
• 35936 Tweets
Mahomes
• 29667 Tweets
Real Facts
• 29465 Tweets
जम्मू कश्मीर
• 28691 Tweets
Asaramji Bapu Case
• 25813 Tweets
Kelce
• 25311 Tweets
Glock
• 24440 Tweets
物価目標
• 20113 Tweets
Olave
• 13973 Tweets
Pinned Tweet
Inspired by the new o1 model, I hacked together g1, powered by Llama-3.1 on
@GroqInc
. It uses reasoning chains to solve problems.
It solves the Strawberry problem ~70% of the time, with no fine tuning or few shot techniques.
A thread 🧵 (with GitHub repo!)
58
134
795
Whisper is now available on
@Groqinc
. With the combined power of fast audio to text and 1200+ tokens/second speed, I had an idea 💡.
⚡️I built groqnotes to turn any audio, like a youtube video about transformers, into organized notes in less than a minute. Open source!
(A
27
104
761
Hacked together an app today powered by Llama-3.1 on
@GroqInc
. Inspired by o1, it uses reasoning chains to solve problems. Maybe prompting is all you need?
Works great on the Strawberry problem, even triple checks the solution! All within 4 seconds of thinking… at the speed of
35
35
424
Imagine you want to learn about the technology behind LLMs. You instantly get an 100 page book with chapters, content, and structure.
What if you find the language too technical? You can change the prompt and the book – all 100 pages – adapts to your needs.
What if you want
17
72
386
🚨 New *multi-modal* model dropped on
@Groqinc
! Llava v1.5 7b is a visual model that can take images as input.
⚡️Try it now in API or console as “llava-v1.5-7b-4096-preview”!
Developers can now build applications on Groq with all three modalities: image, audio, and text!
12
44
268
I’m thrilled to share I started today
@GroqInc
as an AI Applications Engineer intern for this summer!
I’m excited to be working with
@lifebypixels
,
@GrahamRSteele
,
@KapadiaSoami
, and others.
Looking forward to working with the incredible people at Groq and building amazing
24
10
248
@alexalbert__
Artifacts are going to be a game changer for non-technical LLM users who want to build.
Until now, being able to actually run the code to view the output has been a barrier for people with no code experience.
8
12
206
@alexalbert__
It would be great if Claude knew what libraries were supported so it doesn’t use unsupported libraries when creating the artifact.
8
1
204
@GroqInc
is incredibly fast, currently up to 1200+ tokens/second. But what can you do with that speed? ⚡️
📊 Introducing StockBot, a lightning fast AI chatbot powered by Llama3-70b on Groq that responds with live stock charts, financials, news, and screeners. All open source!
18
27
186
Groqnotes powered by
@GroqInc
and
@streamlit
reached 100 stars on GitHub and 70,000 views on X in less than 24 hours… Thank you so much for your support! Excited to keep building
4
10
157
StockBot powered by
@GroqInc
is now a Vercel Template!
Get started with Groq + tool calling enabling generative UI. All on Next.js and
@vercel
AI SDK.
@GroqInc
is incredibly fast, currently up to 1200+ tokens/second. But what can you do with that speed? ⚡️
📊 Introducing StockBot, a lightning fast AI chatbot powered by Llama3-70b on Groq that responds with live stock charts, financials, news, and screeners. All open source!
18
27
186
6
17
154
Here is the GitHub repo:
6
10
138
Groq is insanely fast. How fast? So fast that it can *write entire books in seconds*.
I built groqbook to do just that. Open source!
(A thread 🧵)
7
25
116
GroqBook running Llama3 on
@GroqInc
now exports styled PDF books from markdown!
Prompt —> Full PDF Book in less than a minute!
Groq is insanely fast. How fast? So fast that it can *write entire books in seconds*.
I built groqbook to do just that. Open source!
(A thread 🧵)
7
25
116
3
21
94
@svpino
I think the counterpoint here is that new developers will have a hard time adding much value to a company if all they can do is basic prompting of AI. Senior developers who have the knowledge to surpass AI abilities in certain tasks will be the most valuable.
5
0
81
Mixture of Agents on
@GroqInc
built by
@KapadiaSoami
allows you to ask questions to a collective intelligence of leading open source models, faster than most closed source models run on their own!
Imagine asking questions to AI, but instead of one model responding, several
Mixture of Agents on Groq
Introducing a fully configurable, Mixture-of-Agents framework powered by
@GroqInc
using
@LangChainAI
You can configure your own MoA version using the
@streamlit
UI through the framework.
details + links below👇🧵
9
131
786
4
16
80
🚨 New *family of models* dropped on
@Groqinc
! Llama-3.2 family including 1b, 3b, 11b, and 90b setting new records.
⚡️Try it now in API or console! Check out the speed…
Llama-3.2-90b better on most benchmarks than GPT-4o-mini!
5
12
70
@alexalbert__
Claude 3.5 is great at improving code. See the first iteration of a tic-tac-toe game and the second, which adds score tracking and improves the UI design as requested.
Iteratively improving output is important with LLMs and Claude 3.5 does it well.
3
3
62
g1 powered by Llama-3.1 70b on
@GroqInc
now has a gradio of app, built by
@_akhaliq
! Find it on the GitHub.
See below example of g1 on gradio using reasoning to solve the Strawberry problem, initially thinking there are 2 ‘R’s, until re-examining and finding 3 ‘R’s.
Inspired by the new o1 model, I hacked together g1, powered by Llama-3.1 on
@GroqInc
. It uses reasoning chains to solve problems.
It solves the Strawberry problem ~70% of the time, with no fine tuning or few shot techniques.
A thread 🧵 (with GitHub repo!)
58
134
795
4
8
59
Wish o1 had access to tools like Web Search and Wolfram Alpha?
Check out what happened when I gave g1 powered by
@GroqInc
access.
💥⚡️In ~8.97 seconds of LLM inference and a few more for web latency, g1 answered a question by:
1. Searching real economic data with Wolfram
Inspired by the new o1 model, I hacked together g1, powered by Llama-3.1 on
@GroqInc
. It uses reasoning chains to solve problems.
It solves the Strawberry problem ~70% of the time, with no fine tuning or few shot techniques.
A thread 🧵 (with GitHub repo!)
58
134
795
6
10
59
GroqNotes powered by
@GroqInc
and
@Streamlit
now has a homepage! Let me know what you think!
Links in reply below:
6
8
55
The xRx framework is a collaboration between newly out of stealth
@8090solutions
and
@GroqInc
.
I believe it is one of the most powerful resources for creating seamless, fast, and intelligent AI apps with multimodal capabilities. And it was just open sourced.
A thread (🧵) with
5
11
53
Did you know the
@GroqInc
docs have a showcase page?
Here are a few examples of the impressive showcased apps powered by Groq:
(A thread 🧵)
4
13
52
GroqNotes powered by
@GroqInc
on
@streamlit
now supports Gemma 2 for content generation!
~30,000 input tokens/second… ⚡️ incredibly fast processing of transcribed audio
The team is on it! Updated to Gemma 2 in GroqCloud and now available to you the developer community. Let us know if this new version of Gemma is useful to your apps. Send pics... 😁
4
8
57
3
9
51
3
12
51
How does it work?
g1 uses reasoning chains, in principle a dynamic Chain of Thought, that allows the LLM to "think" and solve some logical problems that usually otherwise stump leading models.
At each step, the LLM can choose to continue to another reasoning step, or provide a
4
4
47
This is so cool!! Listening to a podcast about my own GitHub repository… AI is making education and knowledge both more accessible and enjoyable.
It’s not perfect, there are a few small errors, but I think that inspires more engagement with the source material.
@BenjaminKlieger
@GroqInc
@sundeep
@MatthewBerman
NotebookLM made you a shot podcast on it, the AI hosts seemed to enjoy it.
Also, your repo is great man, starred a few of your repo's for future use, thank you! (I also had it make another one on your book repo, it was also a fun listen and makes the repo sound great!)
3
6
25
6
4
41
In the past few weeks, we have seen several new state-of-the-art Software Engineering Agents.
Here is a rundown to get caught up:
(A thread 🧵)
2
6
36
🚨Big news! New models released on
@GroqInc
today!
Llama3 8B and 70B both fine tuned for tool use. Great for setting up agentic workflows with Groq that use multiple tools and API calls without significant latency.
Try the new models in the console or in your apps now!
I’ve been leading a secret project for months … and the word is finally out!
🛠️ I'm proud to announce the Llama 3 Groq Tool Use 8B and 70B models 🔥
An open source Tool Use full finetune of Llama 3 that reaches the
#1
position on BFCL beating all other models, including
74
236
1K
2
6
40
How does it perform?
No robust performance testing has been done yet. Since g1 is powered by only one open source model, I imagine leading frontier models like GPT-4o and Claude would still outperform on more complex tasks (although they would likely be slower and more
1
3
36
Thanks to a new PR to GroqNotes by
@untilhamza
, you can now get structured notes from YouTube videos directly!
GroqNotes will download the video, transcribe it, and create organized notes in seconds using Whisper and Llama3 on
@GroqInc
!
1
5
34
StockBot powered by
@GroqInc
passed 140 stars on GitHub over the weekend! Grateful for all your support!
@GroqInc
is incredibly fast, currently up to 1200+ tokens/second. But what can you do with that speed? ⚡️
📊 Introducing StockBot, a lightning fast AI chatbot powered by Llama3-70b on Groq that responds with live stock charts, financials, news, and screeners. All open source!
18
27
186
1
3
32
@bindureddy
I wonder if the speed of token generation impacts scoring as well.
GPT-4o is much faster than GPT-4 and that could make it feel like it has a better response.
4
5
32
"The vision has always been to drive the cost of compute to zero. Groq's goal is to preserve human agency in the age of AI, and we think it needs to be in everyone's hands for that to happen".
Hear
@jonathanross321
interviewed by
@lifebypixels
talk about
@GroqInc
’s mission and
3
4
32
What does this mean?
g1 is not an attempt to fully reproduce o1’s abilities, though I am excited to see what the open source community creates in the coming months! This is an early prototype that, inspired by o1’s output, explores an improved Chain of Thought prompting strategy
2
4
32
Wow! Growing at the speed of
@GroqInc
!
Most exciting, I have already seen five different apps inspired by g1. Always amazed at what people create in open source.
Looking forward to when the community starts training custom models with RL to get closer to o1’s architecture and
Inspired by the new o1 model, I hacked together g1, powered by Llama-3.1 on
@GroqInc
. It uses reasoning chains to solve problems.
It solves the Strawberry problem ~70% of the time, with no fine tuning or few shot techniques.
A thread 🧵 (with GitHub repo!)
58
134
795
1
7
31
GroqNotes is now on
@Replit
! Experience the speed of Whisper on
@GroqInc
transcribing audio and converting to structured notes.
Also new ✨: Choose between using different models on Groq under experimental customizations settings.
Links in reply below:
Whisper is now available on
@Groqinc
. With the combined power of fast audio to text and 1200+ tokens/second speed, I had an idea 💡.
⚡️I built groqnotes to turn any audio, like a youtube video about transformers, into organized notes in less than a minute. Open source!
(A
27
104
761
1
3
30
Big quality improvement in 8B too. See books generated by Llama3-8b versus Llama3.1-8b both running on
@GroqInc
What can you do with Llama quality and Groq speed? You can do Instant. That's what. Try Llama 3.1 8B for instant intelligence on .
125
449
4K
0
6
30
Claude recently launched Artifacts alongside Claude Sonnet 3.5 🤖, creating amazing apps that compile in the browser.
However, it can be challenging for beginners to run React code outside of the Artifact.
💡To solve that problem, I created an open source tutorial to turn
2
3
29
In one day, the amount of stars for StockBot more than doubled… so much excitement about
@GroqInc
+
@vercel
! Feeling energized and grateful!
What do you want to see next from StockBot or other Next.js + Groq apps?
StockBot powered by
@GroqInc
passed 140 stars on GitHub over the weekend! Grateful for all your support!
1
3
32
1
4
28
Groqbook powered by
@GroqInc
just reached 100 GitHub stars! Thank you for the support everyone. I will be releasing a few updates to the codebase soon.
Groq is insanely fast. How fast? So fast that it can *write entire books in seconds*.
I built groqbook to do just that. Open source!
(A thread 🧵)
7
25
116
3
5
29
StockBot is an app that leverages
@GroqInc
’s speed,
@vercel
’s AI SDK, and
@tradingview
’s live widgets to power an AI Chatbot that responds with live, interactive charts tailored to your request.
2
6
27
If you found this helpful, share the thread!
Inspired by the new o1 model, I hacked together g1, powered by Llama-3.1 on
@GroqInc
. It uses reasoning chains to solve problems.
It solves the Strawberry problem ~70% of the time, with no fine tuning or few shot techniques.
A thread 🧵 (with GitHub repo!)
58
134
795
1
1
27
I recently shipped a feature on
@GroqInc
console for easy signup to our weekly developer newsletter!
⚡️Get updates on the latest models released on Groq and see featured developers in the spotlight each week!
Find the signup on the console profile page 👇
1
3
26
Learn how to use LlaVa and Llama Guard on
@GroqInc
! A vision model and moderation model together… both underrated in my opinion!
Dropped a new Groq API Cookbook tutorial! 🧑🍳
Learn how to easily build and deploy an image moderation app with LlaVA and Llama Guard 3 8B powered by
@GroqInc
x
@Gradio
x
@huggingface
.
Try the app:
Learn how to build your own:
11
51
309
2
5
26
Infinite Bookshelf now has a new advanced mode:
✍️ Customize the writing style
🧠 Change the level of complexity
🗂️ Add your own writing as seed content
And configure the models used! All lightning fast running on
@GroqInc
.
Imagine you want to learn about the technology behind LLMs. You instantly get an 100 page book with chapters, content, and structure.
What if you find the language too technical? You can change the prompt and the book – all 100 pages – adapts to your needs.
What if you want
17
72
386
2
4
24
“Ship your apps, databases, and more to production in seconds.”
@Railway
is not exaggerating, it’s a breeze to deploy. Bringing Infinite Bookshelf online with a custom domain took minutes… and that just because I had to create a Dockerfile!
New url:
4
2
24
GroqNotes — Audio or YouTube link to structured notes in ~1 minute using Whisper and Llama3 on
@GroqInc
— has a new look on GitHub!
Groqnotes powered by
@GroqInc
and
@streamlit
reached 100 stars on GitHub and 70,000 views on X in less than 24 hours… Thank you so much for your support! Excited to keep building
4
10
157
4
4
23
The entire project is open source. You can try it hosted here:
And support the project here:
2
4
22
Very cool! Another amazing app powered by
@GroqInc
🔥 Super Fast Adaptive Quiz powered by
@Groq
's Llama 3.1 8B,
@Streamlit
and Educhain ()!
🤯 The quiz gets harder when you're right (and easier when you're wrong)!
Try out the app:
8
19
112
1
6
23
@rohanpaul_ai
The future of pen testing will include testing AI-powered chatbots against databases of jailbreak prompts.
4
0
23
Groq + Replit 🚀
1
2
22
Run Llama 3.1 through
@vercel
AI SDK running on
@GroqInc
!
Swapping models and adding tools require just a few lines of code.
Llama 3.1 + Vercel AI SDK
The most powerful open-source AI model — a few lines of code away.
Learn how to stream text, generated structured data, and create generative UI applications using the new 405B model.
4
27
210
0
1
21
Experience distil-whisper running on
@GroqInc
at 240x speed factor! Unlocking new real time use cases with audio.
We're excited to announce that Distil-Whisper is available at . It is designed to provide faster & less expensive English speech recognition while maintaining accuracy comparable to Whisper Large V3. Read more in our new blog - .
3
13
74
2
5
21
Groqbook reached 500 stars! Thank you to the
@GroqInc
community which has been incredibly supportive.
Developers are clearly very excited about the new things we can build with lighting fast inference ⚡️
Groqbook powered by
@GroqInc
just reached 100 GitHub stars! Thank you for the support everyone. I will be releasing a few updates to the codebase soon.
3
5
29
1
1
20
StockBot is now featured on
@GroqInc
’s showcase page with a full writeup of the vision behind Generative UI.
Imagine a future where AI can create custom, interactive web interfaces at the speed of thought.
Ask about the recent performance of a tech stock and instantly receive
@GroqInc
is incredibly fast, currently up to 1200+ tokens/second. But what can you do with that speed? ⚡️
📊 Introducing StockBot, a lightning fast AI chatbot powered by Llama3-70b on Groq that responds with live stock charts, financials, news, and screeners. All open source!
18
27
186
1
8
18
While building Groqbook 📚, I developed a simple way to track and show
@GroqInc
’s impressive inference statistics.
I just open sourced a new project that teaches you how to integrate this into your app to showcase Groq’s 1200+ tokens/second speed. ⚡️
A thread (🧵)
1
4
18
Check out Gemma 2 9B on
@GroqInc
! Another great open source model at blazing fast speed.
The Groq API is OpenAI endpoint compatible, which means you can swap one of OpenAI’s models with an open source model on Groq by changing just a few lines of code (tutorial link in replies).
Fast to launch & very fast output speed! Groq has launched their Gemma 2 9B offering and is serving it at ~600 output tokens/s
Gemma 2 9B is worthy alternative to Llama 3 8B and other smaller models. It is particularly attractive for generalist and communication-focused
4
22
70
1
5
18
1.5k feels like a milestone… thank you all, never expected the incredible level of support and excitement from the dev community on X.
Looking forward to continuing to ship apps
@GroqInc
and share new projects with you!
2
0
18
Infinite Bookshelf is an improved project originally called GroqBook, announced first here:
Groq is insanely fast. How fast? So fast that it can *write entire books in seconds*.
I built groqbook to do just that. Open source!
(A thread 🧵)
7
25
116
2
2
18
The future of open source AI is very exciting! With lightning fast progress, new use cases open, unlocking realities that were once dreams.
Jonathan Ross, Founder & CEO, Groq: “Open-source wins. Meta is building the foundation of an open ecosystem that rivals the top closed models and at Groq we put them directly into the hands of the developers—a shared value that’s been fundamental at Groq since our beginning. To
10
24
241
2
5
17
@GroqInc
is an amazing place, come join us!
We're hiring full times roles and interns.
Groq Winter Internship program runs from January – May. Applications for Winter 2025 open today.
Join us:
#IJoinedGroq
3
2
38
3
2
16
Groq Chat Streamlit App built by Tony Kipkemboi is a Streamlit app that integrates with Groq to provide a chat interface for fast models powered by Groq.
@Groqinc
in
@streamlit
!
1
2
16
Great video walkthrough of StockBot powered by
@GroqInc
! Thanks for the support
@lifebypixels
Shout out to
@BenjaminKlieger
for bringing me up to speed on his latest LLM demo app for
@GroqInc
called STOCKBOT. Check out this short vid, and if you're a developer he's looking for contributors on Git! Powered by Groq leveraging Vercel AI SDK and
@tradingview
2
1
27
0
2
16
@bertiecharlton
@bindureddy
@sheet_copilot
I’m seeing the same results for a few of my more complex use cases. Will be interesting to see SWE-agent’s performance on SWE-bench when powered with GPT-4o.
Anyone else notice the performance difference between SWE-agent powered by GPT 4 versus Claude 3 Opus?
Seems like a big win for GPT-4 over Opus in AI agent tasks if it carries over to other systems.
(Source: )
1
3
8
1
1
15
Chatgpt has taken the world by storm. Creating essays, reports, even messages, all appearing human-made. In December, at the beginning of winter break, I set out on a mission: Create an intuitive platform to allow anyone to detect AI content. Introducing
1
5
15
StockBot is up to 500 stars and 60 forks!
I added four new GitHub issues as great ways for others to get experience contributing to an open source repository.
Writing code to improve open source apps is how I learned to program full web apps like StockBot.
@GroqInc
is incredibly fast, currently up to 1200+ tokens/second. But what can you do with that speed? ⚡️
📊 Introducing StockBot, a lightning fast AI chatbot powered by Llama3-70b on Groq that responds with live stock charts, financials, news, and screeners. All open source!
18
27
186
2
3
15
Answer Engine built by Developer’s Digest is a Perplexity-like answer engine powered by Groq, Mixtral, Langchain and Brave.
Questions answered with sources, answers, images, videos, and follow-up questions in lightning fast time! ⚡️
1
2
14
Groqnotes then takes the raw transcript and drafts full notes organized into sections with titles:
2
0
13
Thank you
@GroqInc
for the feature! Groq's speed has unlocked a ton of new possibilities for developers.
@BenjaminKlieger
is a developer on the rise and we're humbled by the exciting projects he builds using our LPU™ based solutions via GroqCloud™. His latest build is really useful and a great example showing the potential of tooling with AI. 👀
0
0
9
1
1
14
@omarsar0
Great use case for multi-agent workflows: One LLM performs the task with standard prompting then another LLM call structures the response.
@yang3kc
Solution:
1. Have model give output in unconstrained format
2. Then use a small, cheap model to suck that output into a constrained format
5
3
76
1
2
14
New Google DeepMind paper finds long-context LLMs "consistently outperform RAG in terms of average performance.”
They propose routing between long-context LLMs due to "RAG's significantly lower cost".
As
@GroqInc
continues to innovate and the cost of compute approaches zero, I
Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approach
Google DeepMind compares RAG and long-context LLMs, finding LC outperforms but at a higher cost. Proposes a method to dynamically choose between RAG and LC.
📝
2
34
124
0
3
13
@svpino
That’s a good point. I think we might see new jobs for building software that look very different from today, but I think the principle that lower-level code understanding is valuable will not disappear. We still learn assembly today in university classes despite 99% of
2
0
13
Great app using
@GroqInc
showing how fast inference can be leveraged to utilize complex prompting techniques while maintaining near real-time responses.
Ask a question and in seconds, several agents have discussed and iterated on the response you are provided.
Mixture of Agents on Groq
Introducing a fully configurable, Mixture-of-Agents framework powered by
@GroqInc
using
@LangChainAI
You can configure your own MoA version using the
@streamlit
UI through the framework.
details + links below👇🧵
9
131
786
1
4
13
@HarperSCarroll
featured Infinite Bookshelf powered by
@GroqInc
!
See replies for link to GitHub, its all open source 😄
Here’s how you can use the fastest AI in the world (for LLM inference) for free to write a book about whatever you want. 📖💭🪄
@GroqInc
is amazing
7
14
162
1
1
13
Register for the webinar! It’s time to go beyond the chat interface.
Learn how to build flexible, multimodal apps rendering dynamic UI with xRx.
Join
@8090solutions
& Groq for a webinar unveiling xRx, an innovative open-source framework that empowers developers to create seamless, real-time AI experiences integrating voice, text & visual interactions, powered by Groq LPU technology.
1
2
14
2
3
13
@GroqInc
@vercel
@tradingview
StockBot is integrated with six different
@tradingview
widgets, including:
- Heatmap of Daily Market Performance
- Breakdown of Financial Data for Stocks
- Price History of Stock
- Candlestick Stock Charts for Specific Assets
- Top Stories for Specific Stock
- Stock Screener
2
2
13
Powerful generative UI app! The future of AI creating dynamic interfaces is here…
Meet Meter Command.
A generative UI product for getting information, taking action, and building real-time software – all in natural language.
21
50
612
0
1
13
Magic Spell built by ai-ng and Nick Oates is an AI-powered text editor built with Next.js, Vercel AI SDK and Groq.
Easily integrate AI prompting into your textarea!
1
1
12
@JosttenSackitey
@GroqInc
@lifebypixels
@GrahamRSteele
@KapadiaSoami
Apps that enable new use cases while showcasing Groq’s capabilities. There are some great examples of those here:
0
3
12
@GroqInc
@streamlit
Check out the rest here:
Did you know the
@GroqInc
docs have a showcase page?
Here are a few examples of the impressive showcased apps powered by Groq:
(A thread 🧵)
4
13
52
0
1
12
@alexalbert__
Very helpful! Any prompt for getting Claude to automatically use the cdnjs link for all unsupported libraries?
A full list of included libraries for Artifacts would be helpful!
0
0
12
@lalopenguin
@GroqInc
@sundeep
If there’s interest, I’ll open source! Still playing around with it…. It’s not perfect, but it does a lot better than an LLM out of the box without prompting.
4
0
12
@GroqInc
@vercel
@tradingview
Thanks for reading! To support the project, you can star the repository on GitHub and share this thread!
@GroqInc
is incredibly fast, currently up to 1200+ tokens/second. But what can you do with that speed? ⚡️
📊 Introducing StockBot, a lightning fast AI chatbot powered by Llama3-70b on Groq that responds with live stock charts, financials, news, and screeners. All open source!
18
27
186
1
1
12
As with all generative AI, content may include inaccurate or placeholder information. Groqnotes is in beta and all feedback is welcome!
1
0
10
@vercel
’s AI SDK also makes it simple to set up a leading open source model running on
@GroqInc
in an existing project.
1
5
12
Iterating a solution to complex tasks through several frontier models is incredibly powerful.
You can get a Mixture of Agents effect using Claude Sonnet 3.5 and GPT-4o by passing off a task between the models so they work together to overcome their individual limitations.
Something I've been experimenting with a lot recently is the combination of LLMs for building automated pipelines.
It has become apparent that LLMs have different strengths (i.e., capabilities, understanding, vision/audio/text, language, cost, latency, etc.) so how might we
8
18
163
1
2
12
What is xRx?
xRx is a framework that powers AI applications capable of handling multi-modal input (x), a reasoning engine for the business logic (R, particularly relevant with the recent renewed focus on inference-time reasoning gains!), and multi-modal output (x). Together, xRx
1
1
12
Groqnotes uses Whisper-large on Groq to transcribe audio fast. An example with Google Cloud Tech's video on Transformers:
1
0
11
@GroqInc
@vercel
@tradingview
Groq's speed makes tool calling and providing a response near instantaneous, allowing for a sequence of two API calls with separate specialized prompts to return a response. The future of AI workflows!
1
3
11
Now released!
Inspired by the new o1 model, I hacked together g1, powered by Llama-3.1 on
@GroqInc
. It uses reasoning chains to solve problems.
It solves the Strawberry problem ~70% of the time, with no fine tuning or few shot techniques.
A thread 🧵 (with GitHub repo!)
58
134
795
0
0
11
Innovation
@GroqInc
continues… things just keep getting better for developers!
Groq continues to innovate and has showcased they can achieve >500 tokens/s on Llama 3.1 70B!
@GroqInc
has let us know this has been achieved by:
‣ Improvements in their deterministic scheduling of compute (on the chips) and communication between the chips. This deterministic
4
19
107
1
3
11