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
Prime Intellect Profile picture
@PrimeIntellect
May 7 • 7 tweets • 3 min read • Read on X
Scrolly
The next wave of AI will not be won by better prompts. It will be won by systems that learn from experience.

Today, Prime Intellect Lab is out of beta, open for you to start training your own models.

The era of self-improving agents is here.
Previously, improving a model meant waiting on the frontier labs.

Lab brings the model improvement engine right to you:

Build. Evaluate. Train. Deploy. Image
Lab is launching with self-serve support for models from Nvidia, OpenAI, Meta, Qwen, with more coming soon.

Models range from 1B to 400B parameters covering both dense and MoE architectures, reasoning and non-reasoning modes, and text and image modalities. Image
Under the hood, Lab runs on prime-rl.

We leverage multi-tenant LoRA for async RL and other algorithms, unlocking competitive per-token pricing.

This means you only pay as you go, without worrying about optimizing GPU clusters. Image
Beta users ran 10,000+ training jobs on Lab. They trained agents for browser use, data workflows, and long-horizon coding.

Build your own harness. Turn your data into training tasks. Evaluate your success criteria. Improve your agent. Repeat.

This is what we built Lab for. Image
This same loop is already in prod at leading enterprises.

We help unlock self-improvement flywheels on their data and reward signals from real production tasks.

Lab for Enterprise gives you dedicated capacity, custom environments and deeper support for models at 1T+ scale.Image
Over the next few weeks, we will be sharing more with the community about how Lab unlocks powerful self-improvement for real-world systems.

Start training: app.primeintellect.ai/dashboard/home…

Get in touch: primeintellect.ai/contact

Read more: primeintellect.ai/blog/lab-is-op…

• • •

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

Keep Current with Prime Intellect

Prime Intellect 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 @PrimeIntellect

Prime Intellect Profile picture
Apr 30
Over the past months, Cohort I of our RL Residency has been shipping.

Highlights
- continual learning
- automating AI research (from GPU programming to RL itself)
- embodied environments
- multi-agent systems
- materials science discovery
CARLA-Env – @myainotez

An open-source embodied RL environment based on the CARLA simulator. It provides high-fidelity physics, sensors, and configurable urban scenarios for training and evaluating decision-making agents.

Blog: blog.sinatras.dev/Carla-EnvEnvir…: app.primeintellect.ai/dashboard/envi…
x.com/myainotez/stat…
PMPP-Eval – @myainotez

A dataset and RL environment based on the book “Programming Massively Parallel Processors,” focused on CUDA and GPU programming skills. Includes verifiable coding exercises and a frontier eval based on it.

Blog: blog.sinatras.dev/PMPP-Eval+Jour…

Environment: app.primeintellect.ai/dashboard/envi…

x.com/myainotez/stat…
Read 11 tweets
Prime Intellect Profile picture
Feb 11
Introducing Lab: A full-stack platform for training your own agentic models

Build, evaluate and train on your own environments at scale without managing the underlying infrastructure.

Giving everyone their own frontier AI lab.
We are not inspired by a future where a few labs control the intelligence layer

So we built a platform to give everyone access to the tools of the frontier lab

If you are an AI company, you can now be your own AI lab

If you are an AI engineer, you can now be an AI researcher
Lab unifies everything you need for post-training research into one platform

+ Environments Hub
+ Hosted Evaluations
+ Hosted Training
+ Deployments & Inference

Without needing to worry about the costs of massive GPU clusters or the headaches of low-level algorithm details Image
Read 13 tweets
Prime Intellect Profile picture
Jan 27
We're excited to introduce @arcee_ai's Trinity Large model.

An open 400B parameter Mixture of Experts model, delivering frontier-level performance with only 13B active parameters.

Trained in collaboration between Arcee, Datology and Prime Intellect.
Trinity Architecture

Key design choices:
- Interleaved local + global attention (3:1 pattern)
- Grouped-query + gated attention
- New load-balancing method (SMEBU)
- Depth scaled sandwich norm and QK norm

With extreme sparsity, built for long context and fast inference.Image
Infrastructure

- Large-scale synthetic data generation on ~2k H100s
- Training Trinity Large on 2k B300 GPUs

Training stack:
- Modified torchtitan
- Muon optimizer
- HSDP with FSDP group size 128
- Expert parallelism
- Context parallelism for context extension
- Improvements to recover quickly from hardware failures
Read 7 tweets
Prime Intellect Profile picture
Jan 1
We believe the next breakthrough in long-horizon agents is training models to manage their own context.

Introducing our new research direction on Recursive Language Models.

We are sharing our initial experiments showing the promise of RLMs.

primeintellect.ai/blog/rlm
First introduced by @a1zhang in Oct 2025, the RLM has access to its inputs through a variable in a persistent Python REPL.

The model can inspect & transform that variable with code, and pipe parts of it into sub-LLMs with tools without ever loading the potentially huge input data into its context.Image
RLMs are a simple, flexible form of context folding that doesn't depend on lossy summarization.

Instead, the model proactively delegates context to:

- Python scripts (search, filter, transform)
- Sub-LLMs (fresh instances) for parallel work
- Iterative answer edits until it's actually correct
Read 8 tweets
Prime Intellect Profile picture
Nov 27, 2025
Introducing INTELLECT-3: Scaling RL to a 100B+ MoE model on our end-to-end stack

Achieving state-of-the-art performance for its size across math, code and reasoning

Built using the same tools we put in your hands, from environments & evals, RL frameworks, sandboxes & more
INTELLECT-3 is a 106B parameter Mixture-of-Experts model trained with both SFT and RL on top of the GLM 4.5 Air Base model.

Both stages, including multiple ablations, were carried out on a 512-GPU H200 cluster over the course of two months. Image
Our Training Stack

+ PRIME-RL: Our scalable, asynchronous RL trainer
+ Verifiers: Our unified library used for hundreds of envs and evals on the Environments Hub
+ Sandboxes: Custom container infra optimized for agentic RL
+ Compute: Orchestration & observability for 512 H200s
Read 13 tweets
Prime Intellect Profile picture
Oct 27, 2025
We're scaling our Open-Source Environments Program

As part of this, we're committing hundreds of thousands of $ in bounties and looking for partners who want to join our mission to accelerate open superintelligence

Join us in building the global hub for environments and evals
Over the past 2 months, we've crowdsourced 400+ environments and 80+ verified implementations through our bounties and RL residency across:

+ Autonomous AI Research
+ Browser Automation
+ Theorem Proving
+ Subject-Specific QA
+ Legal/Finance Tasks
+ Many more...
Thank you to everyone whose claimed a bounty or joined the residency!

@alexinexxx @xlr8harder @LatentLich @myainotez @ChaseBrowe32432 @varunneal @vyomdundigalla @amit05prakash @minjunesh @sidbing @unrelated333 @ljt019 @lakshyaag @sid_899 @srthkdev @semiozz @ibnAmjid and more! Image
Read 6 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!

:(