Brbr L4

NVIDIA GPT-OSS 120B

Open chat

Est. 2026 · Signal Lab

Brbr L4Brbr L4

A private chat atelier running on NVIDIA GPT-OSS 120B — 120 billion parameters, streamed to you a token at a time.

120 billion parameters, streaming live

A chat workspace that actually thinks with you.

Brbr L4 runs on NVIDIA's GPT-OSS 120B through the NIM API. Stream answers token by token, hand it files, images or your voice, and let it remember the things worth remembering — all in one fast, keyboard-friendly interface.

Model
gpt-oss-120b
Context
4,096 out
Sign-up
none
brbr l4 · live

Summarise this spec and turn it into a build plan.

L4

Here's the plan, split into three shippable slices — starting with the data model, then the streaming layer.

typescriptcopy
for await (const chunk of stream) {
  answer += chunk.content
}

Everything included

Built for real work, not demos

Key stays server-side

The NVIDIA credential lives only in server environment variables — never in the browser, the UI, or any client request.

Straight to NIM

Every reply is a real call to integrate.api.nvidia.com — no canned responses, no mock layer.

Works on your phone

A responsive layout with a slide-over history panel, thumb-friendly controls and dark or light themes.

Abstract render of accelerated compute hardware
Brbr L4

Open a thread and start talking

No account, no setup. Your chats, memories and preferences stay tied to this browser.

Launch Brbr L4