I had 14 tabs open just to keep up with AI. arXiv, Papers With Code, every leaderboard, HuggingFace, half a dozen RL-env hubs... So I built one screen for all of it. The Bloomberg terminal for AI research. It's called Sophon 🧵

Every model, benchmark, leaderboard, RL env and paper - cross-linked, one screen: • 910 models, full spec sheets • 584 benchmarks with saturation bars • 1.8k RL envs + datasets • 27 leaderboards, ranked live • 34k papers • 372 labs No account. It's all just there.

The part I'm proudest of 👇 Your model has a gap on an eval. Which env or dataset actually closes it? Sophon maps tool → the benchmark it lifts. Pick a capability, get the tools ranked by what they move. Nobody connects "the gap" to "the fix". We do.

See where the frontier really stands: • every leaderboard, live • saturation bars - maxed out, or still climbing? • any models head-to-head as a benchmark matrix Stop guessing who's actually #1.

One feed of what just shipped - models, evals, envs, papers - color-coded, the metric that matters inline. Save the filters you live in as views. ⌘K to jump to anything.

AI moves too fast for 14 tabs. Sophon is one screen for the whole field. Free, no signup. Open the terminal → Tell me what's missing 🙏

ps - a "sophon" is the proton-sized supercomputer the Trisolarans use to watch all of human science in The Three-Body Problem. Felt right.

Most useful page, thanks!

Imba

amazing!

Really amazing work, thank you for making this.

What's the ROI of tracking all of this, though? don't you get lost in so much info?

The attention on the UI🔥, This is amazing man!

The most interesting thing is not just the app but the attention to detail; this UI is a joy to look at, how the heck did you do it? Any skill or resource recommendations to reach these levels? The design really looks premium.

so, soohon when

cool

Nice thing

How is it different from

Looks clean

Like it

LOVE Thisss! I have been searching something similar for so long

this is killer

Just what i need

wow 😳

14 tabs open is a struggle I know too well. Sophon could be a lifesaver for anyone deep into AI research. can't wait to see it in action.

Love the unified view, makes tracking gaps a breeze

14 tabs to one screen. I know that pain, but benchmark aggregation only matters if the metrics map to real tasks. Any embodied AI or robotics envs, or just LLMs?

this is super useful

Love the UI - will try myself. If you want to collab, happy to give you an API that drives - it backfills all arXiv data daily - but doesn’t have the updated benchmark data you have. Looking to find alternative after PwC

3 body problem mentioned :)