Skip to main content
Radar

Every change to the models you run, with its source and its date. Releases, price changes, retirements, API changes, and incidents.Every change to the models you run, with its source.

Follow model changes
BenchLM

ARC-AGI-1 Semi-Private Evaluation (ARC-AGI-1)

We mirror this table; we do not rank on it.

Data verified 36 confirmed releases in the last 30 daysFollow model changes

ARC Prize fluid-intelligence benchmark using novel visual grid transformations.

Benchmark score on ARC-AGI-1 — September 18, 2026

We mirror the published score view for ARC-AGI-1. GPT-6 Astra leads the public snapshot at 98.50%, followed by Claude Opus 5 (97.50%) and Claude Fable 5.1 (97.50%). We do not use these results to rank models overall.

1Proprietary

GPT-6 Astra

OpenAI

98.50%
Overall 82.81Context 1.05M
2Proprietary

Claude Opus 5

Anthropic

97.50%
Overall 81.87
3Proprietary

Claude Fable 5.1

Anthropic

97.50%
Overall 84.74Context 1M

3 modelsReasoningCurrentDisplay onlyUpdated September 18, 2026

Benchmark score table (3 models)

Score
1
GPT-6 AstraOpenAI · Closed
98.50%
2
Claude Opus 5Anthropic · Closed
97.50%
3
Claude Fable 5.1Anthropic · Closed
97.50%

The published ARC-AGI-1 snapshot places GPT-6 Astra first at 98.50%. The third row is 1.00 points behind. The broader top-10 range is 1.00 points, so many of the published results sit in a relatively narrow band.

3 models have been evaluated on ARC-AGI-1. The benchmark falls in the Reasoning category. This category carries a 17% weight in BenchLM.ai's overall scoring system. ARC-AGI-1 is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About ARC-AGI-1

Year

2026

Tasks

Semi-private ARC-AGI-1 evaluation set

Format

Verified accuracy

Difficulty

Abstract visual reasoning

The Opus 5 system card reports the ARC Prize Foundation's verified semi-private max-effort result.

ARC Prize leaderboard

Freshness and provenance

Version

ARC-AGI-1 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does ARC-AGI-1 measure?

ARC Prize fluid-intelligence benchmark using novel visual grid transformations.

Which model scores highest on ARC-AGI-1?

GPT-6 Astra by OpenAI currently leads with a score of 98.50% on ARC-AGI-1.

How many models are evaluated on ARC-AGI-1?

3 AI models have been evaluated on ARC-AGI-1 on BenchLM.

Compare top models on ARC-AGI-1

GPT-6 Astra vs Claude Opus 5Claude Opus 5 vs Claude Fable 5.1
Last updated: September 18, 2026 · BenchLM version ARC-AGI-1 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.