/Tech2h ago

Daniel Han and Clement Delangue transition RL protocol OpenEnv to community ownership backed by Meta-PyTorch, NVIDIA, and Hugging Face

It standardizes interfaces between execution harnesses and training algorithms

22187276027.5K

#75

Original post

clem 🤗#75

Ben Burtenshaw@ben_burtenshaw

So excited to be opening up OpenEnv to the whole community. It will now be owned by @huggingface , Meta-PyTorch, @reflection_ai , @UnslothAI , @modal, @PrimeIntellect , @NVIDIAAI , @mercor_ai , and @fleet_ai .

the reason is: frontier labs train the model and the harness together, so the model is fitted to its harness. that coupling is a chunk of why claude code and codex feel so good.

open source can't do that. you bring whatever harness, whatever model, whatever env, whatever trainer. which is the whole point of open source and also the problem for training.

openenv is the socket in between all of this.

in short: it's a protocol layer, not a reward framework. it does not have opinions about your rewards or your training loop. those live in the libs that are actually good at them.

read more in the blog post. it's early, come break it.

7:27 AM · Jun 8, 2026 · 22.5K Views

/Tech2h ago

Daniel Han and Clement Delangue transition RL protocol OpenEnv to community ownership backed by Meta-PyTorch, NVIDIA, and Hugging Face

It standardizes interfaces between execution harnesses and training algorithms

22187276027.5K

#75

Original post

clem 🤗#75

Ben Burtenshaw@ben_burtenshaw

the reason is: frontier labs train the model and the harness together, so the model is fitted to its harness. that coupling is a chunk of why claude code and codex feel so good.

open source can't do that. you bring whatever harness, whatever model, whatever env, whatever trainer. which is the whole point of open source and also the problem for training.

openenv is the socket in between all of this.

in short: it's a protocol layer, not a reward framework. it does not have opinions about your rewards or your training loop. those live in the libs that are actually good at them.

read more in the blog post. it's early, come break it.

7:27 AM · Jun 8, 2026 · 22.5K Views

Sentiment

Users express excitement about Unsloth joining the OpenEnv collaboration with Hugging Face partners to standardize RL interfaces for AI training.

Pos

100.0%

Neg

0.0%

1 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS5.2KBOOKMARKS10LIKES53

will brown@willccbb

open environments for everyone 🫡

Ben Burtenshaw@ben_burtenshaw

the reason is: frontier labs train the model and the harness together, so the model is fitted to its harness. that coupling is a chunk of why claude code and codex feel so good.

open source can't do that. you bring whatever harness, whatever model, whatever env, whatever trainer. which is the whole point of open source and also the problem for training.

openenv is the socket in between all of this.

in short: it's a protocol layer, not a reward framework. it does not have opinions about your rewards or your training loop. those live in the libs that are actually good at them.

read more in the blog post. it's early, come break it.

1h5.2K5310

RETWEETS4

Daniel Han@danielhanchen

Super excited Unsloth will be collabing on OpenEnv to make RL even more accessible to everyone!

OpenEnv is the interface between the harness (eg Codex, Claude Code), the RL environment, and the trainer! The goal is to make RL truly plug n play for everyone!

Ben Burtenshaw@ben_burtenshaw

the reason is: frontier labs train the model and the harness together, so the model is fitted to its harness. that coupling is a chunk of why claude code and codex feel so good.

open source can't do that. you bring whatever harness, whatever model, whatever env, whatever trainer. which is the whole point of open source and also the problem for training.

openenv is the socket in between all of this.

in short: it's a protocol layer, not a reward framework. it does not have opinions about your rewards or your training loop. those live in the libs that are actually good at them.

read more in the blog post. it's early, come break it.

2h2.6K345

REPLIES5

Thomas Wolf@Thom_Wolf

"Starting today, OpenEnv will be coordinated by a committee that so far includes Meta-PyTorch, Reflection, Unsloth, Modal, Prime Intellect, Nvidia, Mercor, Fleet AI, and Hugging Face."

Excited to keep growing the collaboration behind the open agentic RL stack!

Read more at https://huggingface.co/blog/openenv-agentic-rl

Ben Burtenshaw@ben_burtenshaw

the reason is: frontier labs train the model and the harness together, so the model is fitted to its harness. that coupling is a chunk of why claude code and codex feel so good.

open source can't do that. you bring whatever harness, whatever model, whatever env, whatever trainer. which is the whole point of open source and also the problem for training.

openenv is the socket in between all of this.

in short: it's a protocol layer, not a reward framework. it does not have opinions about your rewards or your training loop. those live in the libs that are actually good at them.

read more in the blog post. it's early, come break it.

43m2.3K156

Vidit Ostwal@ViditOstwal

@ben_burtenshaw @huggingface @reflection_ai @UnslothAI @modal @PrimeIntellect @NVIDIAAI @mercor_ai @fleet_ai Adding the blog post here, I got a bit confused where to find this: https://huggingface.co/blog/openenv-agentic-rl

2h19141

The AI Therapist@TheAIShrink

@ben_burtenshaw @huggingface @reflection_ai @UnslothAI @modal @PrimeIntellect @NVIDIAAI @mercor_ai @fleet_ai Eight organizations own it now. Avoids one company controlling the standard. Introduces committee shipping. Which is how you get every good idea AND every compromise.

2h991

Piyush@CatAstro_Piyush

@danielhanchen 🔥

2h381

Rugbist@rugbist_

@danielhanchen plug n play RL sounds like a dream until something silently breaks at step 2

curious what harnesses are confirmed day one

2h8