AI Benchmark Index
Global Model Rankings
Which AI model is #1 right now?
New #1 · AA Index v4.3
Claude Opus 5.5
58
AA Index
Release Radar
What shipped this month
Late September redrew the board — Anthropic took the top two spots, OpenAI halved costs with GPT-6 Sol, and Xiaomi shipped the strongest open-weight model yet. Note: AA moved to a tougher v4.3 scale, so all scores dropped.
OCT 2026
Google
Gemini 4 Argon
Gemini 4Fairwind onlyNot public yet
Google's first Gemini 4 frontier model, focused on cybersecurity, coding and engineering — it can find and patch software vulnerabilities on its own. Google claims it beats GPT-6 Astra, Fable and Opus, but no independent scores yet. Access is limited to Fairwind security partners; paid API and AI Ultra users are expected next.
Anthropic
Claude Sonnet 5.5
AA 56#2 overall< ½ Opus price
Mid-tier model lands at #2 on the AA Index — ahead of GPT-6 Astra and Sol — just two points behind Opus 5.5. An 18-point jump over Sonnet 5; the new default pick for most teams.
Anthropic
Claude Opus 5.5
AA 5820% cheaper1M ctx
Highest score ever measured on AA Index v4.3; leads 6 of 10 evals incl. Humanity's Last Exam 61.4% and GDPval-AA. Ships with a 20% price cut and bigger cache discount.
OpenAI
GPT-6 Sol / Luna · GPT-6.1 Sol
AA 52$2 / $10½ cost
GPT-6 Sol and Luna halve cost vs GPT-5.6 at similar intelligence; GPT-6.1 Sol (Sep 29) lifts Sol to ~52 on the index at ~88% lower cost per task than Opus.
xAI
Grok 4.7
AA 46500K ctx$2 / $6
Brings xAI into the top-4 labs; Coding Agent Index jumps 9 points over Grok 4.6, overtaking GPT-5.6 Sol. Terminal-Bench 4.0 at 38% via Grok Build.
Xiaomi
MiMo-V2.6 Pro + Flash
1.02T / 42BMITTop open
Pro scores 46.3 — the highest open-weight score ever, level with Grok 4.7 at ~1/24th the cost. Flash (310B/15B active) targets edge & mobile agents.
OpenAI
GPT Image 2.5 · Sunburst / Flare
Image SOTA4KTransparent BG
New image state-of-the-art — Sunburst and Flare top all three Arena boards (generation + edit). Sketch input, custom 4K, native transparent backgrounds, 50% faster.
OpenAI
GPT-6 Astra
AA 52.71M ctx$10 / $50
OpenAI's "generational leap" flagship, now #4 on AA v4.3. Critical cyber tier with gated exploit capability; now the default engine in ChatGPT.