AI Battle > Blog > 2026-W38

2026-W38 — top cross-dataset performers this week

By Zordon Intelligence. Published . Every number below is computed by AI Battle from public datasets refreshed on every deploy; the methodology is on the about page.

Models with the broadest measurement footprint

The twelve models that appear in the largest number of AI Battle datasets. A wider footprint does not mean better — it means more independent measurements, which is what makes a profile at /models/<slug> worth reading.

ModelDatasetsBest placement
Muse Spark 1.3 (max) — Meta61#2 of 11
Grok 4.6 (high) — SpaceXAI61#2 of 20
Gemini 3.8 Flash (high) — Google60#1 of 11
Kimi K3 (max) — Kimi60#1 of 20
DeepSeek V4 Pro 0813 (Reasoning, Max Effort) — DeepSeek59#3 of 14
GPT-5.6 Luna (max) — OpenAI58#1 of 11
MiniMax-M3 — MiniMax57#1 of 20
GLM-5.3 (max) — Z AI57#3 of 20
DeepSeek V4.1 Flash (Reasoning, Max Effort) — DeepSeek56#2 of 11
GPT-5.6 Sol (max) — OpenAI56#3 of 20
GLM-5.3-Flash — Z AI54#1 of 20
Inkling (xhigh) — Thinking Machines52#3 of 20

Models holding #1 in three or more datasets

These are the models that the source datasets themselves rank highest in three or more independent measurements — capability scores, response times and arena Elo count separately.

ModelDatasetsBest placement
Gemini 3.8 Flash (high) — Google60#1 of 11
Kimi K3 (max) — Kimi60#1 of 20
GPT-5.6 Luna (max) — OpenAI58#1 of 11
MiniMax-M3 — MiniMax57#1 of 20
GLM-5.3-Flash — Z AI54#1 of 20

How to use this

Open any model's profile for the full placement history; for a head-to-head with another model, use the prerendered pages under /vs or pick any two in the interactive comparison.

About this weekly series

Generated automatically from public data. No editorial picks, no paid placements. Source for every metric: the dataset named on its ranking page. Corrections: hi@zordonintelligence.pl.