AnthropicUnited States· released Nov 2025

Claude Opus 4.5

Name: Claude Opus 4.5
Author: Anthropic

Ranked #27, 37.6% behind the frontier on average.

Anthropic, United States. One of 12 models from this lab on the Index.

Third-party publishedMedium confidence (7+ comparable tests, 3+ Index categories)

Snapshot September 8, 2026

OpenCharts Index

62.4

Ranked #27 (21 to 30) of 36 · 37.6% behind the frontier on average · 21 comparable tests · 3 counting categories

Drop any single test and the Index lands between 57.4 and 66.4, anywhere from #21 to #30. Neighbours inside that range are ties.

8 of 8 hardest tests · share of the frontier

Distance to today's frontier on the hardest tests. Not a measure of intelligence.

What it's known for

Claude Opus 4.5, in 21 results.

Strongest in agentic (78.9) and reasoning (67.4). It trails the frontier most in math, 59.0% behind on average. No comparable results yet in multimodal and long context. Between 35% and 40% behind the best published results, on average. Drop any single test and it would sit anywhere from #21 to #30.

Strongest category

Agentic

Long-horizon tasks with tools: terminals, desktops, browsers, whole jobs.

78.9

DeepResearch Bench54.8% · 99% of best
GDPval45.5% · 92% of best
Cybench82.0% · 88% of best
METR Time Horizon4.9 h · 82% of best
Terminal-Bench63.1% · 75% of best
Vending-Bench 2$4,967 · 74% of best
APEX-Agents20.7% · 44% of best

Mean of 7 of the category's 8 comparable tests. Counts toward the Index.

Best single result

DeepResearch Bench

Write a PhD-level research brief with web access, graded on coverage, insight and citations.

54.8%

99.1% of the best (55.3%) · #3 of 19.

A research agent lives or dies on whether its sources are real and its synthesis is right.

Where it trails

FrontierMath Tier 4

Solve the hardest FrontierMath tier: problems that take expert mathematicians days.

4.9%

95.0% behind the best published result (97.6%) · #44 of 51.

The deepest end of the set. Progress here signals genuinely new capability.

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.

Multimodal
Long context

Covered, but not averaged into the Index:

Knowledgesingle-test category, shown beside the Index · 60.4
Coding3 of 7 tests · needs 4 · mean 63.4
Human preferencesingle-test category, shown beside the Index · 90.2

Every score, with its source

See it at work

What Claude Opus 4.5 was asked to do, and how it did.

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.

Human preference · mean 90.2 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it

Text ArenaLeaderboard

Answer the same prompt as a rival model; real users vote blind on which answer they prefer.

1,473

45% expected win rate against the board leader (1,507)
#27 of 92

20251101-high-32k · board 2026-09-02

Source: Arena

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

Claude Opus 4.5 against whoever you pick.

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 4.5

#27·Index 62.4C

0 : 11

wins · 0 ties · 11 shared

GPT-6 Astra

#1·Index 99.0A+

GPT-6 Astra comes out ahead on 11 of the 11 tests both have taken.

Category means

Reasoning
67.4
99.6
Knowledge
60.4
100.0
Coding
63.4
97.7
Math
41.0
99.8
Agentic
78.9
98.5
Human preference
90.2
—

Shared tests · biggest gaps first

FrontierMath Tier 4Math
4.9%
97.6%
−95.0
Code Arena (WebDev)Coding
1,495
1,797
−70.2
FrontierMath (Tiers 1–3)Math
34.4%
93.7%
−63.3
ProofBenchMath
36.0%
99.0%
−63.0
ARC-AGI-2Reasoning
37.6%
95.0%
−60.4
APEX-AgentsAgentic
20.7%
46.7%
−54.8
SimpleQA VerifiedKnowledge
45.7%
75.6%
−39.6
WeirdMLCoding
63.7%
92.9%
−31.4
ARC-AGI-1Reasoning
80.0%
98.5%
−18.8
Mock AIME 2024–2025Math
86.1%
100.0%
−13.9
GPQA DiamondReasoning
86.0%
95.8%
−13.7

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 Astra

Scores

Every test, every source

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

Reasoning67.4 · 5/6
Knowledgebeside60.4 · 1/1
Coding63.4 · 3/7
Math41.0 · 4/5
Agentic78.9 · 7/8
Multimodal—
Human preferencebeside90.2 · 1/1
Long context—

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.

GPQA DiamondEpoch-runReasoning

86.0%

86.3% of the best (95.8%), above chance
36th percentile of 78 · n = 198

20251101 · 32k

Claude Opus 4.5

Claude Opus 4.5, in 21 results.

What Claude Opus 4.5 was asked to do, and how it did.

Claude Opus 4.5 against whoever you pick.

Every test, every source

Compare with its neighbours.

The best models, ranked here, working inside Theo.