Vision ArenaLeaderboard
Answer prompts that include an image; real users vote blind between two models' answers.
1,313
best published result
of 54 models
board 2026-08-27
Source: ArenaAnthropicUnited States· released Jun 2026
Claude Fable 5
Best published result on 6 of the 19 tests it has taken.
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
92.7
Ranked #3 (3 to 4) of 36 · 7.3% behind the frontier on average · 19 comparable tests · 3 counting categories
Drop any single test and the Index lands between 91.9 and 95.6, anywhere from #3 to #4. Neighbours inside that range are ties.
Needs more results (3 of 8 so far)
Averaging 93.1 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 Fable 5, in 19 results.
Strongest in math (95.1) and reasoning (93.7). It holds the best published result on ARC-AGI-1, FrontierCode and Mock AIME 2024–2025 and 3 more. It trails the frontier most in coding, 10.6% 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.
Best published result onARC-AGI-198.5%FrontierCode53.5%Mock AIME 2024–2025100.0%SimpleBench81.9%Text Arena1,507Vision Arena1,313
Strongest category
Math
Competition and research mathematics, graded on the final answer or the proof.
95.1
- Mock AIME 2024–2025100.0% · best
- ProofBench95.0% · 95% of best
- FrontierMath (Tiers 1–3)87.0% · 93% of best
- FrontierMath Tier 490.2% · 92% of best
Mean of 4 of the category's 5 comparable tests. Counts toward the Index.
Best single result
ARC-AGI-1
Solve the original Abstraction and Reasoning Corpus: small grids, a handful of examples, one hidden rule.
98.5%
Best published result of 51 models.
The first ARC set is close to saturated at the frontier, so it now shows whether smaller models can generalize at all.
Where it trails
Code Arena (WebDev)
Build a web app from the same prompt as a rival model; real users vote blind on the result.
1,629
44.9 points short of parity with the board leader (1,797) · #6 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.
- Long context
Covered, but not averaged into the Index:
- Knowledgesingle-test category, shown beside the Index · 93.5
- Agentic2 of 8 tests · needs 4 · mean 86.6
- Multimodalsingle-test category, shown beside the Index · 100.0
- Human preferencesingle-test category, shown beside the Index · 100.0
See it at work
What Claude Fable 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.
Multimodal · mean 100.0 across 1 comparable test · a single-test category, shown beside the Index and never averaged into it
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 Fable 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 Fable 5
#3·Index 92.7A
1 : 9
wins · 4 ties · 14 shared
GPT-6 Astra
#1·Index 99.0A+
GPT-6 Astra comes out ahead on 9 of the 14 tests both have taken.
Category means
Reasoning
93.799.6Knowledge
93.5100.0Coding
89.497.7Math
95.199.8Agentic
86.698.5Multimodal
100.0—Human preference
100.0—
Shared tests · biggest gaps first
- Code Arena (WebDev)Coding−44.91,6291,797
- GPQA DiamondReasoning−14.085.9%95.8%
- CritPtReasoning−9.728.6%31.7%
- −7.690.2%97.6%
- −7.187.0%93.7%
- SimpleQA VerifiedKnowledge−6.570.7%75.6%
- ARC-AGI-2Reasoning−6.189.2%95.0%
- SciCodeCoding+6.060.2%56.5%
- ProofBenchMath−4.095.0%99.0%
- APEX-AgentsAgentic−3.645.0%46.7%
- WeirdMLCodingtie91.9%92.9%
- FrontierCodeCodingtie53.5%53.3%
- ARC-AGI-1Reasoningtie98.5%98.5%
- tie100.0%100.0%
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
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
- Reasoning93.7 · 5/6
- Knowledgebeside93.5 · 1/1
- Coding89.4 · 5/7
- Math95.1 · 4/5
- Agentic86.6 · 2/8
- Multimodalbeside100.0 · 1/1
- Human preferencebeside100.0 · 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
85.9%
86.0% of the best (95.8%), above chance
34th percentile of 78 · n = 198
max
Source: Epoch AI — AI Benchmarking HubARC-AGI-2LeaderboardReasoning
89.2%
93.9% of the best (95.0%)
92th percentile of 51 · n = 120
max
Source: Epoch AI — AI Benchmarking HubARC-AGI-1LeaderboardReasoning
98.5%
best published result
of 51 models
xhigh
Source: Epoch AI — AI Benchmarking HubSimpleBenchLeaderboardReasoning
81.9%
best published result
of 46 models
max
Source: Epoch AI — AI Benchmarking HubCritPtLeaderboardReasoning
28.6%
88.5% of the best (32.3%)
91th percentile of 78
max
Source: Epoch AI — AI Benchmarking HubSimpleQA VerifiedEpoch-runKnowledge
70.7%
93.5% of the best (75.6%)
95th percentile of 57 · n = 1,000
xhigh
Source: Epoch AI — AI Benchmarking HubSciCodeLeaderboardCoding
60.2%
97.0% of the best (62.0%)
99th percentile of 73 · n = 338
max
Source: Epoch AI — AI Benchmarking HubWeirdMLLeaderboardCoding
91.9%
99.0% of the best (92.9%)
97th percentile of 69
max
Source: Epoch AI — AI Benchmarking HubFrontierCodeLeaderboardCoding
53.5%
best published result
of 27 models
harness: claude-code · reasoning effort: xhigh
Source: Epoch AI — AI Benchmarking HubCursorBenchLeaderboardCoding
70.5%
96.0% of the best (73.4%)
90th percentile of 21
max · reasoning level: Max
Source: Epoch AI — AI Benchmarking HubCode Arena (WebDev)LeaderboardCoding
1,629
28% expected win rate against the board leader (1,797)
94th percentile of 86
board 2026-09-05
Source: Arena (LMArena) — Leaderboard DatasetMock AIME 2024–2025Epoch-runMath
100.0%
best published result
of 78 models
high
Source: Epoch AI — AI Benchmarking HubFrontierMath (Tiers 1–3)Epoch-runMath
87.0%
92.9% of the best (93.7%)
93th percentile of 61 · n = 290
max
Source: Epoch AI — AI Benchmarking HubFrontierMath Tier 4Epoch-runMath
90.2%
92.4% of the best (97.6%)
98th percentile of 51 · n = 48
max
Source: Epoch AI — AI Benchmarking HubProofBenchLeaderboardMath
95.0%
95.0% of the best (100.0%)
95th percentile of 61
max · reasoning effort: max
Source: Epoch AI — AI Benchmarking HubAPEX-AgentsLeaderboardAgentic
45.0%
94.9% of the best (47.4%)
96th percentile of 49
Vending-Bench 2LeaderboardAgentic
$5,680
78.2% of the best ($11,182) on a log scale
69th percentile of 53
high
Source: Epoch AI — AI Benchmarking HubVision ArenaLeaderboardMultimodal
1,313
best published result
of 54 models
board 2026-08-27
Source: Arena (LMArena) — Leaderboard DatasetText ArenaLeaderboardHuman preference
1,507
best published result
of 92 models
board 2026-09-02
Source: Arena (LMArena) — Leaderboard DatasetNearby on the leaderboard
Compare with its neighbours.
The best models, ranked here, working inside Theo.
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.