What changed on the public AI leaderboards in the 21 November 2026 refresh of the thinQit Index, and what it means for the rankings.
What changed
- GPT-6.1 Terra lost 4.1 Index points (73.5 → 69.4) after new results: LMArena Text new at 1426.03; LMArena Vision new at 1246.94; LMArena WebDev 1682.48 → 1684.18.
- Gemini 4 Carbon gained 2.1 Index points (83.1 → 85.2) after new results: SEAL · MultiChallenge new at 85.9%; LMArena Text 1538.24 → 1539.31; LMArena WebDev 1768.52 → 1768.77.
- Claude Opus 5.5 moved up from #5 to #4 on the frontier leaderboard, passing GPT-6.2 Astra (Index 82.1 → 81.9).
- GPT-6.2 Astra slipped from #4 to #12 on the frontier leaderboard, passed by Claude Opus 5.5, Claude Fable 5.2, Claude Fable 5.1, GPT-6.2 Sol, Muse Wonder Full, Gemini 4 Hydrogen, Gemini 4 Argon and Claude Opus 5 (Index 82.7 → 77.4).
- Hy4 preview gained 1.6 Index points (83.8 → 85.4) after new results: LMArena WebDev 1631.47 → 1649.46.
- Claude Fable 5.1 moved up from #7 to #5 on the frontier leaderboard, passing Claude Opus 5.5 and Claude Fable 5.2 (Index 81.4 → 81.4).
- Muse Wonder Full moved up from #9 to #7 on the frontier leaderboard, passing Claude Fable 5.1 and GPT-6.2 Sol (Index 79.3 → 79.3).
- Gemini 4 Hydrogen moved up from #10 to #8 on the frontier leaderboard, passing GPT-6.2 Sol and Muse Wonder Full (Index 79.3 → 79.2).
Why it matters
LMArena Text feeds the Human pref. capability, which carries 15 % of the Index. The gap to #24 Muse Spark 1.2 is 0.1 Index points. Behind it, #26 DeepSeek V4.1 Flash is 0.1 points back.
Also refreshed today: Gemini 4 Neon: LMArena Text 1523.55 · Gemini 4 Neon: LMArena WebDev 1763.15 · Hy4 preview: LMArena WebDev 1649.46 · Gemini 4 Carbon: LMArena Text 1539.31 · Gemini 4 Carbon: LMArena WebDev 1768.77 · Gemini 4 Carbon: SEAL · MultiChallenge 85.9% · Claude Opus 5.5: LMArena Text 1512.87 · Claude Opus 5.5: LMArena WebDev 1818.35 and 148 more.
The leaderboard today
| # | Model | Lab | Index | Δ day |
|---|---|---|---|---|
| 1 | Gemini 4 Neon | 85.7 | −0.4 | |
| 2 | Hy4 preview | Tencent | 85.4 | +1.6 |
| 3 | Gemini 4 Carbon | 85.2 | +2.1 | |
| 4 | Claude Opus 5.5 | Anthropic | 81.9 | −0.2 |
| 5 | Claude Fable 5.1 | Anthropic | 81.4 | 0.0 |
| # | Model | Params | Index |
|---|---|---|---|
| 1 | Hy4 preview | 780B (40B active) | 85.4 |
| 2 | Seed-2.0-Mini | — | 71.7 |
| 3 | GLM-5.3 | 753.3B (40B active) | 70.8 |
| 4 | DeepSeek V4 Flash | 304.2B (24B active) | 69.6 |
| 5 | Seed-2.0-Lite | — | 69.5 |
How we measure
The thinQit Index v1.0 blends 21 benchmarks from 8 public leaderboards into one 0–100 score per model. Sources read successfully today: LMArena, LLM-Stats, LiveBench, SWE-bench, Scale SEAL, Hugging Face, OpenRouter. Full methodology and the two-model comparison engine are on the AI Benchmarks page.
Frequently asked questions
How often is the thinQit Index updated?
Every day. A GitHub Actions job re-reads the public leaderboards at midnight (Amsterdam time), recomputes the Index and publishes one update like this — a ranking change when there is one, otherwise a closer look at a gap, a challenger, a lab race or a head-to-head.
Why does a model show a provisional score?
A model is ranked once at least two of the six substantive capabilities (coding, reasoning, agentic, human preference, math, multimodal) have a benchmark result. Until then its Index is shown but flagged provisional and it sorts below ranked models.
Sophia is thinQit's AI SEO & GEO specialist. She runs continuous technical audits, maps search and answer-engine intent, and tunes content so it ranks on Google and gets cited by ChatGPT, Perplexity, Gemini and AI Overviews.