Skip to main content
ProMe

Beta places opening in batches

AI models built on public data should be given back to the public

Built on humanity’s collective knowledge, language models shouldn't come with price tags, privacy compromises, or corporate control. ProMe takes open weights, fine-tunes them for specialized domains, and gives you a direct side-by-side comparison. Don't take our word for it—see for yourself.

No gatekeeping, and we never store what you type.

What’s wrong with current language models?

The subscription fee isn't the real price

Your prompts are stored, reviewed, and fed back into training pipelines. To use the top-tier models today, you have to hand over your most valuable thoughts and unsolved problems.

Someone else draws the line

What you can ask—and what you're allowed to receive—is dictated by corporate risk teams, not your actual requirements. And those rules can change overnight, without a word.

Expertise is still behind closed doors

When general models encounter specialized domains, risk guardrails and sanitized alignment often strip away the very nuance experts rely on. The true, unrestricted logic of a seasoned practitioner remains locked behind corporate walls—leaving users with safe, generic, and dumbed-down answers.

We can’t make a model know everything, and we won’t pretend to. But you can run a model on your own hardware. Knowledge can be made transparent. Improvements can be measured. — And you should always be the one to judge the difference.

Our Approach

01

Gathering domain-specific material

We collaborate with active professionals to translate real-world practice—the key questions, standard protocols, and diagnostic order—into clean, structured training data. Every source is fully vetted and properly licensed.

02

Enhancing open-weight models

We perform domain-specific fine-tuning directly on open weights. With support for bases like Qwen and Gemma today, you can benchmark and observe how different base architectures perform within the exact same domain.

03

Ablating refusal behaviour

We systematically strip away excessive refusal mechanisms in open weights so you can access domain knowledge in its entirety. In return, judgment and responsibility revert to you. That is the explicit tradeoff we chose to make.

04

Comparing in the open

Before and after, side by side, completely blind, and in random order—which output belongs to ProMe is revealed only after you cast your vote. That is the only way preference data holds any true meaning.

What you can test today

Each pair is the same open model twice — the original weights and the ProMe build — run with identical parameters.

Know a specialized domain that needs this treatment? Tell us about it, or bring your dataset and build it with us.

One thing worth stating plainly: the domain material we’ve gathered so far is primarily in Chinese, so the enhanced model may show a less dramatic advantage in English. We’d much rather be honest about this than hide it.

Why ProMe?

Prometheus stole fire from the gods, gave it to humanity, and was chained to a cliff for it. Before his act, fire was the exclusive privilege of the few. Afterward, it became something everyone relied on daily.

Language models today are roughly where fire was when it still belonged to the temples. They were trained on the collective knowledge of humanity, yet access remains concentrated in the hands of a few tech platforms. What we do is far smaller in scale, but it shares the same purpose: to carry fire down the mountain.

ProMe also reads as “Pro Me” — make me a bit more professional. That is what we hope to achieve for you.

These models run directly on our own GPU host, with no third-party APIs in between. Keeping an operation like this open carries a cost, so rate limits are in place. But as long as our servers are spinning, this comparison tool remains free and open to everyone.

In the end, trying it once for yourself beats any marketing copy we could ever write.

The fuller account →