AnthropicUnited States· released Jul 2026
Best published result on 2 of the 20 tests it has taken.
Anthropic, United States. One of 12 models from this lab on the Index.
Available in Theo as Theo Code
Theo orchestrates it for the steps it does best and always shows which engine answered.
Snapshot September 8, 2026
OpenCharts Index
92.6
Ranked #4 (3 to 4) of 36 · 7.4% behind the frontier on average · 20 comparable tests · 3 counting categories
Drop any single test and the Index lands between 91.7 and 94.3, anywhere from #3 to #4. Neighbours inside that range are ties.
Needs more results (3 of 8 so far)
Averaging 87.2 on the 3 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 reasoning (96.0) and math (91.1). It holds the best published result on OSWorld 2.0 and Vending-Bench 2. It trails the frontier most in coding, 9.4% behind on average. No comparable result yet in long context. Between 5% and 10% behind the best published results, on average. Drop any single test and it would sit anywhere from #3 to #4.
Strongest category
Reasoning
Hard, novel problems: graduate science, abstract puzzles, expert exams.
96.0
Mean of 5 of the category's 6 comparable tests. Counts toward the Index.
Best single result
Complete a multi-hour workflow across real desktop and web applications, end to end, with the environment changing as you work.
31.4%
Best published result of 9 models.
Computer use is how an agent reaches every tool that has no API, and long workflows are where agents lose the thread.
Where it trails
Build a web app from the same prompt as a rival model; real users vote blind on the result.
1,688
30.5 points short of parity with the board leader (1,797) · #3 of 86.
People judging finished apps side by side is the most honest measure of front-end quality.
Not measured yet
No comparable result yet in this category. 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.
Agentic · mean 97.3 across 3 comparable tests · not counted toward the Index yet (fewer than half the category's tests)
Complete a multi-hour workflow across real desktop and web applications, end to end, with the environment changing as you work.
31.4%
best published result
of 9 models
max
A separate, far harder benchmark from OSWorld-Verified: binary completion is low for every model, and the leaderboard also reports a partial-credit score.
Source: XLANG LabRun a simulated vending business for a year: ordering, pricing and cash flow. The score is the final balance.
$11,182
best published result
of 53 models
Andon Labs estimates a strong human operator at roughly $63,000, so every model is far from the ceiling.
Source: Andon LabsComplete professional-services work in consulting, law and finance as an agent, graded against expert rubrics.
43.5%
91.8% of the best (47.4%)
#4 of 49
max
Source: MercorEvery 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 Opus 5
#4·Index 92.6A
0 : 8
wins · 6 ties · 14 shared
GPT-6 Astra
#1·Index 99.0A+
GPT-6 Astra comes out ahead on 8 of the 14 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.
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.
93.9%
97.3% of the best (95.8%), above chance
92th percentile of 78 · n = 198
max
Source: Epoch AI — AI Benchmarking Hub90.4%
95.2% of the best (95.0%)
96th percentile of 51 · n = 120
max
Source: Epoch AI — AI Benchmarking Hub97.5%
99.0% of the best (98.5%)
90th percentile of 51 · n = 100
max
Source: Epoch AI — AI Benchmarking Hub80.6%
98.4% of the best (81.9%)
98th percentile of 46
29.1%
90.2% of the best (32.3%)
92th percentile of 78
max
Source: Epoch AI — AI Benchmarking Hub59.9%
79.2% of the best (75.6%)
80th percentile of 57 · n = 1,000
max
Source: Epoch AI — AI Benchmarking Hub55.7%
89.7% of the best (62.0%)
83th percentile of 73 · n = 338
max
Source: Epoch AI — AI Benchmarking Hub91.8%
98.8% of the best (92.9%)
96th percentile of 69
max
Source: Epoch AI — AI Benchmarking Hub53.4%
99.8% of the best (53.5%)
96th percentile of 27
max · harness: claude-code · reasoning effort: medium
Source: Epoch AI — AI Benchmarking Hub70.0%
95.4% of the best (73.4%)
85th percentile of 21
max · reasoning level: Max
Source: Epoch AI — AI Benchmarking Hub1,688
35% expected win rate against the board leader (1,797)
98th percentile of 86
max · board 2026-09-05
Source: Arena (LMArena) — Leaderboard Dataset98.9%
98.9% of the best (100.0%)
91th percentile of 78 · n = 45
max
Source: Epoch AI — AI Benchmarking Hub85.6%
91.4% of the best (93.7%)
90th percentile of 61 · n = 290
max
Source: Epoch AI — AI Benchmarking Hub73.2%
75.0% of the best (97.6%)
88th percentile of 51 · n = 48
max
Source: Epoch AI — AI Benchmarking Hub99.0%
99.0% of the best (100.0%)
97th percentile of 61
max
Source: Epoch AI — AI Benchmarking Hub31.4%
best published result
of 9 models
max
Source: Epoch AI — AI Benchmarking Hub43.5%
91.8% of the best (47.4%)
94th percentile of 49
max
Source: Epoch AI — AI Benchmarking Hub$11,182
best published result
of 53 models
1,290
47% expected win rate against the board leader (1,313)
89th percentile of 54
high · board 2026-08-27
Source: Arena (LMArena) — Leaderboard Dataset1,493
48% expected win rate against the board leader (1,507)
93th percentile of 92
high · board 2026-09-02
Source: Arena (LMArena) — Leaderboard DatasetNearby on 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.