LLMBenchmark.ai

Enter opens the highlighted result. Esc closes.

About LLMBenchmark.ai

LLMBenchmark.ai is an LLM benchmark leaderboard that tracks 356 AI models across 190 public benchmarks, with API prices, updated daily.

Who is LLMBenchmark.ai?

LLMBenchmark.ai is built for developers and teams who pick a model for real work and need current scores and API prices side by side, each with its source.

LLMBenchmark.ai was launched in October 2026 by Take Group, a Polish software company that runs on AI agents. The team behind it has used large language models in daily work since ChatGPT came out in November 2022.

Who runs LLMBenchmark.ai

Jake Slugocki

Co-founder, R&D at Take Group

Jake Slugocki (Jakub Sługocki in Polish) runs LLMBenchmark.ai. He co-founded Take Group in 2019 with Bartłomiej Krawczyk and leads its research and development: AI agents, internal tools and new products.

Jake Slugocki is not a programmer by training. He builds Take Group's internal apps himself with Claude Code and Codex, and he designs the AI agents that run most of the company's daily work.

Jake Slugocki has used large language models every day since ChatGPT launched in November 2022. He tries every new model on real tasks, from frontier models like Claude, GPT and Gemini to small open-weight models, and keeps the ones that earn a place in an agent.

LLMBenchmark.ai grew out of that habit. Choosing the model for each agent means checking current scores and prices across many boards, so the site puts them in one place and shares what Take Group checks anyway.

Jake Slugocki on LinkedIn

Jake Slugocki, co-founder of Take Group
Jake Slugocki, Co-founder, R&D at Take Group

How Take Group works with AI

Take Group runs on AI agents. These are the facts behind that sentence, as of October 2026.

  1. 15 people run more than 5,000 websites. Jake Slugocki estimates that without agents the same work would take a team of about 100.

  2. Every internal tool at Take Group has its own MCP server, so one agent can reach any part of the company and repeat any process or task.

  3. New models are tried on real agent work, not only on demos. The scores on LLMBenchmark.ai are the first filter.

About Take Group

Take Group is a Polish software company that builds AI tools that scale work, and it runs on its own AI agents. Jake Slugocki and Bartłomiej Krawczyk founded it in Katowice in 2019.

Take Group also ships its own AI products. Cotlet lets AI agents such as Claude Code, Codex or Cursor work in apps that have no API or MCP server.

Where the numbers come from

LLMBenchmark.ai reads every score from the board that published it, across the 18 public sources below. Each number on the site links to its source and shows the day it was read. A value nobody published is marked as missing, never estimated.

Source Results read Last read
OpenRouter 356
Artificial Analysis (via OpenRouter) 213
Epoch AI benchmarking hub 198
Artificial Analysis 195
OpenRouter (provider speed) 175
LMArena (Hugging Face dataset) 169
Hugging Face 151
Design Arena 141
Vals AI boards 132
ALE-Bench (Sakana AI) 104
LMArena WebDev 101
MMLU Pro (Vals AI) 92
GPQA (Vals AI) 91
AIME (Vals AI) 54
LiveBench 43
SWE-bench (swebench.com) 38
Aider polyglot leaderboard 22
Vals Index 20

What LLMBenchmark.ai adds

LLMBenchmark.ai computes two numbers of its own. The Vibe Coding Index is a weighted average of four public coding benchmarks. Cost per solved task is the price of one SWE-bench Verified attempt divided by the share of tasks a model resolves.

How often the data changes

The LLMBenchmark.ai dataset is rebuilt from its sources every day. The current data is from October 20, 2026.

Use the data

The LLMBenchmark.ai data is licensed CC BY 4.0: use it, commercially too, and credit LLMBenchmark.ai with a link. Embedded charts and tables keep a visible source link. The full dataset is available as JSON, and agents can query it through a free API and MCP server.

Company

LLMBenchmark.ai is run by TAKE GROUP sp. z o.o., ul. Wincentego Janasa 4, 40-855 Katowice, Poland. KRS 0000871398, NIP (tax ID) 6342984273.