Sam Altman says OpenAI next open-source model will be much better than people expect It will run incredibly powerful models locally and surprise people with what’s possible "i think you all will be astonished"

Running top-tier models locally changes everything. Privacy, speed, cost, all get better. This will bring more interesting product ideas that weren’t viable before.

I hope the upcoming OpenAI open source model has 1. Permissible license e.g. MIT or Apache (just like DeepSeek and Qwen), but I am not holding my breath here. 2. Comes in varieties e.g. family of models: 4B, 8B, 14B and 32B. 3. Has great multi-lingual support, like Gemma 3 or Qwen3. 4. Has excellent prompt and system prompt adherence 5. Has at least 128k input token window, plus can output at least 8k tokens. I am not expecting SOTA level in coding or maths or having as good niche knowledge as the larger models, that's not just realistic.

If Sam’s hyping it this much, either it’s genuinely next-level or he’s setting the bar insanely high. Let’s see if they deliver

Advanced reasoning doesn’t take huge models. Pretty soon our toaster will be smarter than us.

All talk, no action. Release GPT-5 already man

Damn this would really tear the market open

If we can run frontier-level models on a laptop, every backyard robot and community DAO just got a brain upgrade. Super keen to see how OpenAI squeezes that much juice into local compute. Already thinking up token-driven neighborhood AI co-ops.

I watched it live and couldn't stand the heat.

I just want deliverance man. Only Sama and Elon are two guys who have more model announcements than model releases

This changes everything! You won’t just chat with AI, you’ll build with it, remix it, embed it into real workflows. Local inference could be the real distribution play.