Home/About

What is CodeSOTA?

CodeSOTA is an independent platform that tracks state-of-the-art results in machine learning. Think of it as a scoreboard for AI models across every task that matters.

The Problem
1
Scattered information
Papers, blogs, Twitter, leaderboards everywhere
2
Outdated benchmarks
Papers with Code shut down. Other sources stale.
3
Unverified claims
Companies self-report. No independent verification.
4
No code links
Finding implementations is a treasure hunt.
CodeSOTA Solution
1
One source of truth
286+ benchmarks across 17 research areas
2
Updated weekly
New models added as they're released
3
Verified results
We run benchmarks. Source links on every entry.
4
Links to code
GitHub, HuggingFace, API docs for every model

The Story: From Papers with Code to CodeSOTA

How we got here

2018Papers with Code launches

First centralized ML benchmark tracking

2019Meta acquires PWC

Becomes part of Facebook AI

2023PWC updates slow

Less active maintenance

July 2025PWC shuts down

Meta discontinues the platform

Aug 2025CodeSOTA launches

Independent, verified benchmarks

17 Research Areas, One Platform

Click an area to see what's inside

Click an area above to explore

How to Use CodeSOTA

From question to decision in 5 steps

You have a question
Which model is best for document OCR?
?
"Which model should I use for extracting text from scanned documents?"

What a Benchmark Looks Like

Example: MMLU (Massive Multitask Language Understanding)

Leaderboard
Rank
Model
Score
#1
GPT-5API
92.3%
#2
Claude 3.5 OpusAPI
91.8%
#3
Gemini Ultra 2API
90.5%
#4
Llama 4 405BOpen
88.7%
What You Get
Verified Scores
From papers or our own runs
Source Links
Click to verify any result
Implementation Links
GitHub, HuggingFace, APIs

CodeSOTA by the Numbers

All data freely available as JSON

286+
Benchmark Results
17
Research Areas
143
Models Tracked
86
Datasets

Who Uses CodeSOTA?

Researchers

Track SOTA for your papers. Compare your model to baselines. Find prior work.

Engineers

Pick the right model for production. Find open-source alternatives. Get implementation code.

Decision Makers

Understand the AI landscape. Make informed build vs buy decisions. Track competition.

Ready to Explore?

Start browsing benchmarks or explore AI building blocks to find the right model for your task.

Browse BenchmarksAI Building Blocks