AnthropicUnited States· released Oct 2025
Provisional: 15 comparable tests so far. Ranked once it reaches 6 across 3 counting categories.
Anthropic, United States. One of 12 models from this lab on the Index.
Available in Theo as Theo Fast
Theo orchestrates it for the steps it does best and always shows which engine answered.
Snapshot September 8, 2026
OpenCharts Index
32.7
Provisional: not enough comparable tests or counting categories to rank yet · 15 comparable tests · 2 counting categories
Drop any single test and the Index lands between 26.7 and 37.6.
Needs more results (2 of 8 so far)
Averaging 23.1 on the 2 it has; the score opens at 4 of 8.
Distance to today's frontier on the hardest tests. Not a measure of intelligence.
What it's known for
Strongest in agentic (35.8) and reasoning (29.5). It trails the frontier most in reasoning, 70.5% behind on average. No comparable results yet in multimodal and long context. Below 40% of the best published results, on average.
Strongest category
Agentic
Long-horizon tasks with tools: terminals, desktops, browsers, whole jobs.
35.8
Mean of 4 of the category's 8 comparable tests. Counts toward the Index.
Best single result
Solve the hardest tier of the MATH competition dataset to an exact final answer.
96.4%
98.2% of the best (98.1%) · #5 of 6.
Multi-step algebra and geometry, graded only on the final answer.
Where it trails
Run a simulated vending business for a year: ordering, pricing and cash flow. The score is the final balance.
$459
100.0% behind the best published result ($11,182) on a log scale · #44 of 53.
Long-horizon decisions whose consequences compound. It exposes models that drift or forget.
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.
Math · mean 82.5 across 2 comparable tests · not counted toward the Index yet (fewer than half the category's tests)
Solve the hardest tier of the MATH competition dataset to an exact final answer.
96.4%
98.2% of the best (98.1%)
#5 of 6
20251001 · 32k
Epoch flags likely training contamination on this set; treat small gaps as meaningless.
Source: Epoch AI (internal runs)Solve competition mathematics problems that have a single exact integer answer.
66.7%
66.7% of the best (100.0%)
#76 of 78 · n = 45
20251001 · 32k
45 problems: one problem is 2.2 points, so gaps of a few points are within noise.
Source: Epoch AI (internal runs)Every 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.
Claude Haiku 4.5
provisional·Index 32.7F
0 : 10
wins · 0 ties · 10 shared
GPT-6 Astra
#1·Index 99.0A+
GPT-6 Astra comes out ahead on 10 of the 10 tests both have taken.
Category means
Reasoning
Knowledge
Coding
Math
Agentic
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.
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.
71.2%
65.3% of the best (95.8%), above chance
3th percentile of 78 · n = 198
20251001 · 32k
Source: Epoch AI — AI Benchmarking Hub4.0%
4.2% of the best (95.0%)
2th percentile of 51 · n = 120
20251001 · 32k
Source: Epoch AI — AI Benchmarking Hub47.7%
48.4% of the best (98.5%)
6th percentile of 51 · n = 100
20251001 · 32k
Source: Epoch AI — AI Benchmarking Hub0.0%
0.0% of the best (32.3%)
0th percentile of 78
20251001
Source: Epoch AI — AI Benchmarking Hub13.2%
17.5% of the best (75.6%)
5th percentile of 57 · n = 1,000
20251001
Source: Epoch AI — AI Benchmarking Hub43.3%
69.8% of the best (62.0%)
31th percentile of 73 · n = 338
20251001
Source: Epoch AI — AI Benchmarking Hub45.4%
48.9% of the best (92.9%)
21th percentile of 69
20251001
Source: Epoch AI — AI Benchmarking Hub1,329
6% expected win rate against the board leader (1,797)
14th percentile of 86
20251001 · board 2026-09-05
Source: Arena (LMArena) — Leaderboard Dataset66.7%
66.7% of the best (100.0%)
3th percentile of 78 · n = 45
20251001 · 32k
Source: Epoch AI — AI Benchmarking Hub96.4%
98.2% of the best (98.1%)
20th percentile of 6
20251001 · 32k
Source: Epoch AI — AI Benchmarking Hub35.5%
41.9% of the best (84.7%)
31th percentile of 40
20251001 · agent: Goose
Source: Epoch AI — AI Benchmarking Hub8.9%
18.8% of the best (47.4%)
17th percentile of 49
20251001
Source: Epoch AI — AI Benchmarking Hub$459
0.0% of the best ($11,182) on a log scale
0th percentile of 53
20251001
Source: Epoch AI — AI Benchmarking Hub45.5%
82.3% of the best (55.3%)
33th percentile of 19 · n = 100
20251001 · low
Source: Epoch AI — AI Benchmarking Hub1,413
37% expected win rate against the board leader (1,507)
15th percentile of 92
20251001 · board 2026-09-02
Source: Arena (LMArena) — Leaderboard DatasetTop of the leaderboard
28 of the models on this page run inside Theo today. Theo picks the right one for each step and always shows which engine answered.