Project benchmark

ARC-AGI-1 · Abstract reasoning

ARC Prize · 289 configurations with published scores. Review each metric and its test conditions separately.

Report identity & conditions

Data as of2026-11-03· Report published—. See individual records for run dates and test conditions.

Leaderboard updated2026-10-28T16:04:23.365Z

Run configurations remain separate, including different reasoning settings.
Individual test dates are not given. Update dates and model release dates from the report remain separate and do not substitute for actual test dates.
Costs follow the report’s run scope and dollar amounts; they are not model API unit prices.
ARC-AGI-1 · Semi-private evaluation set. Benchmark versions are compared separately.
Task score on the semi-private evaluation set, as a percentage. It is not ranked across benchmark versions or evaluation scopes.
Cost per task as reported, in USD/task. Total costs and task counts are not invented. This is neither cost per successful task nor a current API quote.
Human reference results are shown separately and excluded from model/system rankings and cost comparisons.

Read the original report ↗

Left metricRight metric

Compare metrics within the same report. Each chart explains its direction and scope.
See the details for full values and run conditions.

Task score

Task score on the semi-private evaluation set, as a percentage. It is not ranked across benchmark versions or evaluation scopes.: showing 8 of 289 records with values. See details for configurations and dates.

Chart data · 8 entries

Name	%
Claude Fable 5 (Max)	98.5
Claude Fable 5 (XHigh)	98.5
Claude Opus 5.5 (High)	98.5
Gemini 3.8 Flash (High)	98.5
Gemini 4 Argon (High)	98.5
Gemini 4 Argon (Medium)	98.5
GPT-6 Astra (High)	98.5
GPT-6 Astra (XHigh)	98.5

Cost per task

Cost per task as reported, in USD/task. Total costs and task counts are not invented. This is neither cost per successful task nor a current API quote.: showing 8 of 289 records with values. See details for configurations and dates.

Chart data · 8 entries

Name	USD / Task
GPT-5 Nano (Minimal)	0.002
GPT-4.1-Nano	0.002
GPT-6 Luna (None)	0.002
DeepSeek V4 Flash 0731 (None)	0.002
Qwen3-235b-a22b Instruct (25/07)	0.003
GPT-6 Luna (Low)	0.003
GPT-5 Nano (Low)	0.003
Nemotron 3.5 Super VL (None)	0.004

Human reference

Human baselines are shown separately and are not ranked alongside models and systems.

Stem Grad

Task score: 98%
Cost per task: 10 USD / Task

Human Panel

Task score: 98%
Cost per task: 17 USD / Task

Full results & model versions

289 / 289configurations have published scores in this scope. Scored configurations are listed; missing values remain “Not provided.” Different run configurations of the same model are listed separately.

Claude Haiku 4.5 (Thinking 16K)

Model reported by the benchmark

As reportedOrganization: Anthropic

claude-haiku-4-5-20251001 · CoT

Release time reported by the benchmark:2025-10-01

Task score: 37.33%
Cost per task: 0.1042 USD / Task

Grok 4.20 (Reasoning)

Model reported by the benchmark

As reportedOrganization: xAI

grok-4.20-beta-0309b-reasoning · CoT

Release time reported by the benchmark:2026-03-09

Task score: 89.5%
Cost per task: 0.35 USD / Task

Claude Fable 5.1 (Low)

Model reported by the benchmark

As reportedOrganization: Anthropic

anthropic-claude-fable-5-1 · CoT

Release time reported by the benchmark:2026-09-01

Task score: 90%
Cost per task: 0.366044 USD / Task

Grok 3 Mini (Low)

Model reported by the benchmark

As reportedOrganization: xAI

CoT

Release time reported by the benchmark:2025-02-19

Task score: 16.5%
Cost per task: 0.0099 USD / Task

Grok 4 (Thinking)

Model reported by the benchmark

As reportedOrganization: xAI

CoT

Release time reported by the benchmark:2025-07-09

Task score: 66.67%
Cost per task: 1.0136 USD / Task

Magistral Medium

Model reported by the benchmark

As reportedOrganization: Mistral

CoT

Release time reported by the benchmark:2025-06-10

Task score: 5.91%
Cost per task: 0.1015 USD / Task

Claude Opus 4 (Thinking 8K)

Model reported by the benchmark

As reportedOrganization: Anthropic

claude-opus-4-20250514-thinking · CoT

Release time reported by the benchmark:2025-05-14

Task score: 30.67%
Cost per task: 0.7408 USD / Task

Nemotron 3.5 Super VL (None)

Model reported by the benchmark

As reportedOrganization: NVIDIA

nvidia-nemotron-3-5-super-vl-120b-a12b · CoT

Release time reported by the benchmark:2026-10-15

Task score: 2%
Cost per task: 0.0038 USD / Task

Claude Fable 5 (High)

Model reported by the benchmark

As reportedOrganization: Anthropic

anthropic-claude-fable-5 · CoT

Release time reported by the benchmark:2026-06-09

Task score: 95.5%
Cost per task: 0.766625 USD / Task

GPT-5 Mini (Minimal)

Model reported by the benchmark

As reportedOrganization: OpenAI

CoT

Release time reported by the benchmark:2025-08-07

Task score: 5.33%
Cost per task: 0.0057 USD / Task

Grok 4.7 (XHigh)

Model reported by the benchmark

As reportedOrganization: xAI

xai-grok-4-7 · CoT

Release time reported by the benchmark:2026-09-21

Task score: 89.5%
Cost per task: 0.77845 USD / Task

Claude Sonnet 4.5 (Thinking 32K)

Model reported by the benchmark

As reportedOrganization: Anthropic

claude-sonnet-4-5-20250929 · CoT

Release time reported by the benchmark:2025-09-29

Task score: 63.67%
Cost per task: 0.5193 USD / Task

Claude Sonnet 5.5 (XHigh)

Model reported by the benchmark

As reportedOrganization: Anthropic

anthropic-claude-sonnet-5-5 · CoT

Release time reported by the benchmark:2026-09-28

Task score: 96.5%
Cost per task: 0.143341 USD / Task

Claude Haiku 4.5 (Thinking 32K)

Model reported by the benchmark

As reportedOrganization: Anthropic

claude-haiku-4-5-20251001 · CoT

Release time reported by the benchmark:2025-10-01

Task score: 47.67%
Cost per task: 0.2558 USD / Task

o3-mini (High)

Model reported by the benchmark

As reportedOrganization: OpenAI

o3-mini · CoT

Release time reported by the benchmark:2025-01-31

Task score: 34.5%
Cost per task: 0.3989 USD / Task

DeepSeek V4 Pro 0813 (None)

Model reported by the benchmark

As reportedOrganization: DeepSeek

deepseek-v4-pro-0813 · CoT

Release time reported by the benchmark:2026-08-13

Task score: 13%
Cost per task: 0.023428 USD / Task

GPT-4.1

Model reported by the benchmark

As reportedOrganization: OpenAI

gpt-4.1 · Base LLM

Release time reported by the benchmark:2025-04-14

Task score: 5.5%
Cost per task: 0.039 USD / Task

Deepseek R1

Model reported by the benchmark

As reportedOrganization: DeepSeek

CoT

Release time reported by the benchmark:2025-01-22

Task score: 15.8%
Cost per task: 0.06 USD / Task

GPT-5.6 Terra (High)

Model reported by the benchmark

As reportedOrganization: OpenAI

openai-gpt-5-6-terra · CoT

Release time reported by the benchmark:2026-07-09

Task score: 92%
Cost per task: 0.19 USD / Task

DeepSeek V4 Pro 0813 (Max)

Model reported by the benchmark

As reportedOrganization: DeepSeek

deepseek-v4-pro-0813 · CoT

Release time reported by the benchmark:2026-08-13

Task score: 90%
Cost per task: 0.296178 USD / Task