MetaUnited States· released Oct 2026· open weights
Ranked #25 early, tested so far in 2 of the 5 categories the Index averages.
Meta, United States. One of 7 models from this lab on the Index.
Snapshot November 10, 2026
OpenCharts Index
63.3
Ranked #25 (16 to 38) of 80, early · 36.7% behind the frontier on average · 11 comparable tests · 2 counting categories
Part of this Index is an estimate. On the 2 categories it is tested in, it averages 69.0. The rest is filled in from models tested in every category, because it has too few results to count there yet: Coding, Agentic, Multimodal. Such estimates have been within 5.8 points about nine times in ten.
Its band runs from 56.5 to 71.4: where the Index could land once it is tested in every category, or if any single test were dropped. That covers #16 to #38, and neighbours inside that range are ties.
Needs more results (3 of 5 so far)
Averaging 55.2 on the 3 it has; the score opens at 4 of 5.
Distance to today's frontier on the hardest tests. Not a measure of intelligence.
What it's known for
Strongest in reasoning (70.3) and math (67.7). It trails the frontier most in math, 32.3% behind on average. No comparable results yet in multimodal and human preference. Between 35% and 40% behind the best published results, on average. Part of its Index is an estimate. It has too few results in coding, agentic and multimodal for those categories to count, so they are filled in from models tested in every category. On the categories that do count it averages 69.0. Across its band it could sit anywhere from #16 to #38.
Strongest category
Reasoning
Hard, novel problems: graduate science, abstract puzzles, expert exams.
70.3
Mean of 4 of the category's 8 comparable tests. Counts toward the Index.
Best single result
Solve competition mathematics problems that have a single exact integer answer.
98.1%
98.1% of the best (100.0%) · #28 of 98.
Olympiad-style problems need long, careful chains of reasoning, and there is no partial credit.
Where it trails
Build a web app from the same prompt as a rival model; real users vote blind on the result.
1,526
68.5 points short of parity with the board leader (1,818) · #47 of 110.
People judging finished apps side by side is the most honest measure of front-end quality.
Not measured yet
No comparable result yet in these 2 categories. A blank is a blank, never a zero, and it does not lower the Index.
Covered, but not averaged into the Index:
See it at work
Every test it has taken, by category, with the kind of task it faced, who produced the number and where the result landed against the best published one. The examples are original and illustrative, never items from the datasets themselves.
Reasoning · mean 70.3 across 4 comparable tests · counts toward the Index
Solve the original Abstraction and Reasoning Corpus: small grids, a handful of examples, one hidden rule.
89.0%
90.4% of the best (98.5%)
#34 of 67 · n = 100
xhigh
Close to saturated at the frontier; scores depend on the compute budget a lab chose.
Source: ARC Prize FoundationAnswer a PhD-level multiple-choice question in physics, chemistry or biology that a web search will not settle.
88.6%
89.0% of the best (96.4%), above chance
#53 of 96 · n = 198
xhigh
Frontier models cluster in the eighties and nineties; Epoch's standard errors on this test are 2 to 3 points.
Source: Epoch AI (internal runs)Infer the hidden rule from a few input and output grid pairs, then apply it to a new grid.
57.5%
60.5% of the best (95.0%)
#39 of 67 · n = 120
xhigh
Scores depend on the compute budget a lab chose; ARC Prize publishes cost per task beside every score and this ranking does not.
Source: ARC Prize FoundationSolve an unpublished research-level physics problem to a numeric or symbolic answer checked by an official server.
14.9%
41.3% of the best (36.0%)
#45 of 102
xhigh
Epoch AI's review rated version 1 Flawed (October 2026): two audits found scoring defects in at least a fifth of its problems. The leaderboard moved to a revised version 2 on October 22, 2026, and Epoch AI's file does not say which version each score comes from.
Source: CritPt / Artificial AnalysisEvery result is read against the best published one on its test: percent scores as a share of the best above chance, Arena ratings as the expected win rate against the board leader, open-ended values on a log scale. Sitting a harder exam never lowers a model. Every number links to the publisher that produced it.
Head to head
Choose a rival. Every counted test both have taken appears side by side as a share of the best published result on that test, with the category means above. A win is a higher share; a gap under 2 points is a tie, because one item on a 45-problem exam is 2.2 points and publishers' own standard errors are of that order.
Muse Spark Open
#25·Index 63.3C
1 : 9
wins · 1 tie · 11 shared
GPT-6 Astra
#1·Index 93.7A
GPT-6 Astra comes out ahead on 9 of the 11 tests both have taken.
Category means
Reasoning
Knowledge
Coding
Math
Agentic
Multimodal
Human preference
Shared tests · biggest gaps first
Bars are distances to the best published result on each test (percent scores above chance, Arena ratings as win rate against the leader, open-ended values on a log scale). Raw values as the publishers report them.
Open GPT-6 AstraScores
The best published run per benchmark, the raw value as the publisher reports it, who produced the number, how close it comes to the best published result on that test, and where the model sits among every model scored on it. Percent scores are read above chance, Arena ratings as a win rate against the board leader, open-ended values on a log scale. A test with too few models to compare against is shown but not counted, and so is an archived test: one whose publisher no longer covers new models.
Category means
Mean · tests taken of the category's comparable tests. A faded category either rests on fewer than half of them or is a single-test category (marked “beside”), which is shown next to the Index and never averaged into it.
88.6%
89.0% of the best (96.4%), above chance
45th percentile of 96 · n = 198
xhigh
Source: Epoch AI — AI Benchmarking Hub57.5%
60.5% of the best (95.0%)
42th percentile of 67 · n = 120
xhigh
Source: Epoch AI — AI Benchmarking Hub89.0%
90.4% of the best (98.5%)
47th percentile of 67 · n = 100
xhigh
Source: Epoch AI — AI Benchmarking Hub14.9%
41.3% of the best (36.0%)
56th percentile of 102
xhigh
Source: Epoch AI — AI Benchmarking Hub56.6%
60.2% of the best (94.0%)
72th percentile of 75 · n = 1,000
xhigh
Source: Epoch AI — AI Benchmarking Hub58.4%
87.4% of the best (66.9%)
83th percentile of 96 · n = 338
xhigh
Source: Epoch AI — AI Benchmarking Hub1,526
16% expected win rate against the board leader (1,818)
58th percentile of 110
max · board 2026-11-09
Source: Arena (LMArena) — Leaderboard Dataset98.1%
98.1% of the best (100.0%)
71th percentile of 98 · n = 45
xhigh
Source: Epoch AI — AI Benchmarking Hub69.8%
73.4% of the best (95.1%)
61th percentile of 78 · n = 290
xhigh
Source: Epoch AI — AI Benchmarking Hub31.7%
31.7% of the best (100.0%)
49th percentile of 68 · n = 48
xhigh
Source: Epoch AI — AI Benchmarking Hub33.4%
39.7% of the best (84.2%)
12th percentile of 58
Nearby on the leaderboard
29 of the models on this page run inside Theo today. Theo picks the right one for each step and always shows which engine answered.