@vikingmute: Who created this amazing website? https://sophon.at It collects and displays all AI-related information and content: papers, newest models, benchmarks, leaderboards. Papers can be viewed online directly, very comprehensive. Also has a feed to subscribe for the latest news. And this…
Summary
This article recommends a website called Sophon, which aggregates AI papers, models, benchmarks, leaderboards, and reinforcement learning environments. It provides real-time rankings, comparisons, and subscription features, and is hailed as the Bloomberg terminal for AI research.
View Cached Full Text
Cached at: 06/05/26, 07:08 AM
Which god made this website? https://t.co/qM6FJuo28V It collects and displays all information and content related to AI: papers, latest models, benchmarks, leaderboards. Papers can even be viewed online directly — it’s extremely comprehensive.
There’s also a feed, so you can subscribe and stay up to date with the latest news.
Also, the word “Sophon” turns out to be the meaning of “智子” (a concept from the Three-Body Problem). https://t.co/pUP4Ig4s9L
Sophon - AI evals, model leaderboards & RL environments
Source: https://sophon.at/homepage Sophon
The Bloomberg terminal for AI research
Every model, benchmark, leaderboard, RL environment and paper on one screen, with the signals that show what is rising and what moves a score.
Open the terminal (https://sophon.at/)Browse the feed (https://sophon.at/feed)
906models
599benchmarks
1.8kRL envs
27leaderboards
39kpapers
373labs
1Claude Opus 4.8 2GPT-5.2 3Gemini 3 Ultra 4Llama 4 405B ### Live standings Every leaderboard, ranked in real time. (https://sophon.at/leaderboards)Opus GPT-5 Gem 3 MMLU 92 90 88 GSM8K 95 96 94 SWE 71 65 68 ### Head to head Models side by side as a benchmark matrix. (https://sophon.at/compare)reasoningmathcodeagents verifiers-mathlifts7 swe-gymlifts5 ### What lifts a score Envs and datasets ranked by the benchmarks they move. (https://sophon.at/recommender)AllRL envsFrontier+ modelGemini 3 Ultra$5/M2h evalAIME 2025top 78%5h toolterminal-bench-rllifts 41d paperScaling RL envs2.3k★1d ### What just shipped One typed stream. Save the filters you live in as views. (https://sophon.at/feed)swe⌘K SWE-bench Verified Claude Opus 4.8 prime-intellect / math Arena Elo ### Find anything ⌘K across models, evals, tools, papers and labs. (https://sophon.at/)1M Context $5/M Price T·I·A Modality MMLU GPQA ### Spec sheets Context, price, modality and full benchmark coverage. (https://sophon.at/models)Scaling RL environments Self-rewarding LMs Process reward models ### Your watchlist Star papers, models and evals to track. Browser-local, no account. (https://sophon.at/collections)trained with averifiable reward signalacross envs ### Highlight the source Mark up a paper’s PDF in four colors. Pin a note to any passage. (https://sophon.at/papers)Notes Reproduces the SWE-bench claim. Compare reward shaping against v2. Only stored in your browser. ### Private notes Jot a thought on any paper. Stored only in your browser. (https://sophon.at/papers)
Plugged into the RL-env ecosystem
Browse all (https://sophon.at/tools)
Links-first, no lock-in.298 env-to-benchmark lift links.
Harbor
Laude
Prime Intellect
1.3k envs
OpenReward
352 envs
Atropos
111 envs
OpenEnv
34 envs
Open the terminal
Enter Sophon (https://sophon.at/)
Similar Articles
@VincentLogic: Drowning in new Arxiv papers every day? Head spinning. Just discovered a treasure trove of a website that aggregates the latest AI papers and model benchmarks. Clean interface, just check Trending or filter by week/month. Best part: each paper directly links to the benchmarks and models it uses.
Recommend a free website sophon.at/papers that aggregates the latest AI papers and model benchmarks. Clean interface, supports Trending or weekly/monthly filtering. Each paper directly links to its benchmarks and models.
@jinchenma_ai: The best compilation of high-quality AI sources on the web, save it now! You can throw this article to Codex + Obsidian, let AI compile an index directory, then later when you want AI to search for quality information, you can let it search according to this directory.
Recommend an article that compiles the best high-quality AI sources on the web, and suggest using Codex and Obsidian to compile an index directory for future AI search of quality information.
@VincentLogic: Discovered a powerful tool for high-quality AI information sources! follow-builders is an open-source project that tracks the daily updates of top AI figures worldwide and automatically summarizes and pushes them to you. Created by Zhang Ziya (who transitioned from a humanities background at Harvard to AI), its core philosophy is solid: "Follow builders, not influencers." It ignores trend-chasing influencers and focuses on real AI product builders...
Introduces follow-builders, an open-source project that automatically tracks updates from AI builders and generates summary notifications, aiming to help users access high-quality information.
@Jackywxsz: https://x.com/Jackywxsz/status/2070743217721770316
The article recommends 5 high-quality AI information sources (AI Hot, BestBlogs.dev, AIGC Weekly, TW93 Blog, The Road to AGI), and shares the author's information acquisition logic based on attention management, emphasizing depth rather than chasing trends.
@XAMTO_AI: Brothers and sisters, don't say it's hard to find information gaps anymore! This magical tool scans 10 platforms at once, and directly condenses what the whole internet has been talking about and betting on in the past 30 days into one report! last30days-skill has topped GitHub Trending #1! 7000+ stars in a week, 13k total stars! Not...
An AI agent-driven search engine that aggregates the latest content from platforms like Reddit, X, YouTube, etc., and automatically generates cross-validated reports. It has topped GitHub Trending #1.