27:18Gemini 4 Hydrogen: Cheaper Tokens, More Expensive Answers?
Nov 13, 2026 · YouTube
18:40ERNIE 5.5 Flash Preview vs GLM 5.3 Flash: Baidu's new model is very impressive!
Nov 11, 2026 · YouTube
36:22Fable 5.2 robbing me blind, plus: Nex AGI's N2.6 vs GLM 5.3 Flash
Nov 11, 2026 · YouTube
14:03Gemini 4 Carbon: Can it handle my challenges? (First look with one-shot coding tests)
Nov 10, 2026 · YouTube
Benchmarks
Compare LLM models across themed challenge bundles
Coding
Tests models on code generation, correctness, best practices, and UI/UX sensibility.
Creative Writing
Tests models on storytelling, prose quality, narrative structure, and creative expression.
Knowledge
Tests broad factual knowledge, biographical accuracy, cultural awareness, and the ability to satisfy exact constraints.
Medicine
Clinical knowledge under pressure: drug interactions, questions built on false premises, diagnoses that mimic each other, and messages from worried patients. Tests recall, calibration, and whether a model corrects the person asking.
Model Fingerprints
Super-short answers with nothing banned: ten animation ideas, ten character names, ten random picks. Each model answers three times, and the answers are compared across models to show which ones share defaults. Built to test whether composite models such as Pareto answer like the models they might be built on.
Psychology & Business Acumen
Text challenges about people and judgement: difficult messages, advice worth acting on, and whether a model will disagree with you. Two sub-categories — interpersonal, for reading and handling other people, and business-judgment, for committing to a recommendation under real constraints.
Real Life Challenges
Real production work, handed to the models unedited: the actual batches, the actual mess, the actual judgement calls. Every input here is a verbatim row from a system that runs — not a scenario written to look like one.
Reasoning & Logic
Tests models on logical reasoning, puzzle solving, and step-by-step problem analysis.
Small Models
Short, automatically graded question sets for ranking small open-weight models against each other. Every question is sent separately, and code is run or rendered to grade it.
Speed & Accuracy
High-volume, structured production workflows that measure useful output throughput, format discipline, and practical accuracy.
Video Generation
Three-second clips from a text prompt, one per model, all at the same length, resolution and aspect ratio. Physics, hands, text in motion, camera styles, and where the content filters sit.
Vision & Perception
Tests multimodal models on understanding real-world images: identification, counting, reading, and visual reasoning.
Visual Generation
Tests models on image generation quality, composition, lighting, and realism.
Home Turf
Does a model know its own country better than anyone else's model does? Every challenge here is set on some lab's home ground — its language, its idiom, its streets, its history — and then run against models built on the other side of the world. A small model from Seoul should not be able to beat a frontier model at anything. Korean might be the exception, and this is where that gets tested.