Amit Sony

AI Benchmark Index
NOV 2026
LIVE DATA
Global Model Rankings

Which AI model is #1 right now?

New #1 · 5 points clear
Claude Fable 5.2
65
AA Index
Release Radar
What shipped this month
November so far: Anthropic retook #1 with Claude Fable 5.2, Google's Gemini 4 Carbon slipped to #2 after under two weeks on top, Moonshot opened Kimi K3.1 Pro, and Meta added a cheap agent model. All scores use AA Index v5.
NOV 2026
Anthropic
Claude Fable 5.2 · Mythos 5.2
New #1 · Nov 9
AA 65$8 / $4030% faster
Takes #1 on AA Index v5 by 5 points over Gemini 4 Carbon. Big gains in science, knowledge and coding; OSWorld 89.3%. Available to all users, 20% cheaper than Fable 5.1. Mythos 5.2, the less-restricted twin, stays limited to verified security and life-science orgs.
Meta
Muse Ember
New · Nov 9
$0.10 inputText+image+audioAA 36
Meta's lightest, cheapest model, built for high-volume agent work like tool calls, routing and extraction. Designed to pair with Muse Wonder: "Wonder plans, Ember handles the volume."
Google
Gemini 4 Carbon
#2 · Oct 28
AA 601M ctx$2 / $10 promo
Held #1 for under two weeks before Fable 5.2 passed it. Beats Argon on all 17 of Google's tests. Still limited to Fairwind security partners; no public date yet.
OpenAI
GPT Image 2.6 · Sunburst / Flare
Nov 4
ImageGIF + SVG~6s API
Successor to GPT Image 2.5 with animated GIF output, SVG export and API generation in about 6 seconds. Arena scores pending — GPT Image 2.5 keeps #1 on the chart until then.
Moonshot AI
Kimi K3.1 Pro (open weights)
Weights · Nov 4
2.8T / 104BOpenAA 47
Weights now on Hugging Face. The strongest open-weight model, and it leads Terminal-Bench 4.0 (67%) ahead of Fable 5.2 and Opus 5.5.
Artificial Analysis
Intelligence Index v5
Oct 29
New scale60% private2 new evals
Biggest index upgrade yet: most tests now use private data, so models can't be trained on the answers. All scores on this page use v5.
OpenAI
GPT-6.1 Terra
Oct 26
AA 511.1M ctx$1 / $5
Mid-tier refresh that scores 51 on AA v5, above Gemini 4 Argon, at a fifth of Astra's input price. The value pick in OpenAI's line-up.
Anthropic
Claude Opus 5.5
#3 · Sep 22
AA 59Coding #1$10 / task
Now #3 overall, but still #1 on the AA Coding Agent Index (62) and Anthropic's recommended everyday model — at well under half Fable 5.2's cost per task.