LLM Benchmark Snapshot, November 9 2026: Same Leaders, Tighter Races

Our benchmarks page takes a snapshot of the public LLM leaderboards every time the daily runs, and each snapshot keeps its own permanent URL. This is the photo for the week of November 9, 2026, compared against last week's post (November 3).

The headline: a quiet week at the top

Nothing changed hands at number one this week. Gemini 4 Carbon still leads the LMArena text arena and BenchLM, and Claude Opus 5.5 still leads WebDev. The movement happened one or two places below the leaders, and in every case the gaps are small enough to call ties.

Text arena: Gemini first, Claude reshuffles

On the LMArena text leaderboard, Gemini 4 Carbon (High) sits at about 1539 Elo, essentially unchanged from 1540 last week. Gemini 4 Argon (High) holds second at about 1530.

Third place swapped inside the Claude family: Claude Fable 5.1 (Max) is now third at 1512.73 and Claude Opus 5.5 (High) fourth at 1512.71. That is a gap of two hundredths of a point, so read it as a tie. Last week the order was the other way around (Opus 5.5 at about 1516, Fable 5.1 at about 1514).

BenchLM: the same order, a slightly wider lead

BenchLM's overall leaderboard keeps the same top three: Gemini 4 Carbon (88.7, up from 88.4), Claude Opus 5.5 (86.3) and GPT-6 Astra (85.8, down from 86.1). Carbon's lead over second place grew from about 2.1 to about 2.4 points, while the gap between second and third opened from a fraction of a point to about half a point.

WebDev: Claude extends its lead

Claude Opus 5.5 (Max) still leads the LMArena WebDev arena and moved up to about 1818, from about 1810. Behind it the order changed: Claude Sonnet 5.5 (xHigh) climbed to second at about 1773, Gemini 4 Carbon (High) is third at about 1768, and GPT-6 Astra (Max), second last week, dropped to fourth at about 1767. Positions two to four are within seven points of each other.

Aider polyglot is still frozen

The Aider polyglot board is identical for another week: GPT-5 (high) at 88.0, GPT-5 (medium) at 86.7 and o3-pro (high) at 84.9. It has not added recent frontier models in months, so treat it as a historical reference.

Sources disagree, and that is the point

Three leaderboards, three answers: Gemini 4 Carbon on general chat preference and on BenchLM's aggregate, Claude Opus 5.5 on web development, and a GPT-5 entry on a coding board that stopped updating. Pick the leaderboard that matches your task, and check whether it is still being maintained.

Data note: our LiveBench feed has still returned no rows since the October 18 snapshot, so LiveBench is missing from this comparison again.

Explore the data