AnthropicUnited States· released Feb 2026

Claude Sonnet 4.6

Name: Claude Sonnet 4.6
Author: Anthropic

Ranked #22, 36.1% 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)

Available in Theo as Theo Balanced

Theo orchestrates it for the steps it does best and always shows which engine answered.

Use it in Theo

Snapshot September 8, 2026

OpenCharts Index

63.9

Ranked #22 (18 to 28) of 36 · 36.1% behind the frontier on average · 20 comparable tests · 3 counting categories

Drop any single test and the Index lands between 61.0 and 69.7, anywhere from #18 to #28. Neighbours inside that range are ties.

Needs more results (3 of 8 so far)

Averaging 72.2 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

Claude Sonnet 4.6, in 20 results.

Strongest in agentic (65.6) and coding (63.8). It trails the frontier most in reasoning, 37.7% behind on average. No comparable result yet in long context. Between 35% and 40% behind the best published results, on average. Drop any single test and it would sit anywhere from #18 to #28.

Strongest category

Agentic

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

65.6

DeepResearch Bench54.9% · 99% of best
Vending-Bench 2$7,204 · 86% of best
Terminal-Bench53.4% · 63% of best
APEX-Agents23.7% · 50% of best
OSWorld 2.09.3% · 30% of best

Mean of 5 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.

54.9%

99.3% of the best (55.3%) · #2 of 19.

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

Where it trails

CritPt

Solve an unpublished research-level physics problem to a numeric or symbolic answer checked by an official server.

3.1%

90.3% behind the best published result (32.3%) · #50 of 78.

Research physics is far past textbook recall. Only a few models produce anything a physicist would accept.

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.

Long context

Covered, but not averaged into the Index:

Knowledgesingle-test category, shown beside the Index · 47.0
Math2 of 5 tests · needs 3 · mean 65.4
Multimodalsingle-test category, shown beside the Index · 89.2
Human preferencesingle-test category, shown beside the Index · 90.0

Every score, with its source

See it at work

What Claude Sonnet 4.6 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 90.0 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.

1,472

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

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 Sonnet 4.6 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 Sonnet 4.6

#22·Index 63.9C

0 : 12

wins · 0 ties · 12 shared

GPT-6 Astra

#1·Index 99.0A+

GPT-6 Astra comes out ahead on 12 of the 12 tests both have taken.

Category means

Reasoning
62.3
99.6
Knowledge
47.0
100.0
Coding
63.8
97.7
Math
65.4
99.8
Agentic
65.6
98.5
Multimodal
89.2
—
Human preference
90.0
—

Shared tests · biggest gaps first

CritPtReasoning
3.1%
31.7%
−88.5
Code Arena (WebDev)Coding
1,521
1,797
−66.0
FrontierCodeCoding
24.3%
53.3%
−54.1
ProofBenchMath
45.0%
99.0%
−54.0
SimpleQA VerifiedKnowledge
35.5%
75.6%
−53.0
APEX-AgentsAgentic
23.7%
46.7%
−48.5
ARC-AGI-2Reasoning
60.4%
95.0%
−36.4
WeirdMLCoding
66.1%
92.9%
−28.9
SciCodeCoding
46.8%
56.5%
−15.6
Mock AIME 2024–2025Math
85.8%
100.0%
−14.2
ARC-AGI-1Reasoning
86.5%
98.5%
−12.2
GPQA DiamondReasoning
87.4%
95.8%
−11.9

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

Reasoning62.3 · 4/6
Knowledgebeside47.0 · 1/1
Coding63.8 · 6/7
Math65.4 · 2/5
Agentic65.6 · 5/8
Multimodalbeside89.2 · 1/1
Human preferencebeside90.0 · 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

87.4%

88.1% of the best (95.8%), above chance
43th percentile of 78 · n = 198

32k

Claude Sonnet 4.6

Claude Sonnet 4.6, in 20 results.

What Claude Sonnet 4.6 was asked to do, and how it did.

Claude Sonnet 4.6 against whoever you pick.

Every test, every source

Compare with its neighbours.

The best models, ranked here, working inside Theo.