ARC-AGIBrowse 284 families

ARC-AGI

Can the model solve difficult problems that require more than factual recall?

Version / measure
Domain
Reasoning
Ranked models
26
Reported values
66
Reports
3 · lab + maintainer
Direction
Higher is better
Status
Living benchmark

Rankings

Best reported score per model across effort settings

26 models
#ModelBest settingScore
1GPT-6 Astra - Provider AdapterOpenAIhighcost n/a6 values · Sep 24 · source99.95%
2GPT-6 AstraOpenAImaxcost n/a6 values · Sep 24 · source62.71%
3Gemini 3.8 Flash - Provider AdapterGooglehighcost n/a3 values · Sep 24 · source35.00%
4Claude Opus 5Anthropicmaxcost n/a1 value · Jul 24 · source30.20%
5Claude Opus 5Anthropichighcost n/a1 value · Sep 24 · source30.16%
6Gemini 3.8 FlashGooglehighcost n/a3 values · Sep 24 · source10.37%
7GPT-5.6 SolOpenAIreportedcost n/a2 values · Jul 24 · source7.80%
8GPT-5.6 SolOpenAImaxcost n/a5 values · Sep 24 · source7.78%
9Grok 4.6xAIxhighcost n/a1 value · Sep 24 · source2.11%
10Claude Opus 4.8Anthropichighcost n/a1 value · Sep 24 · source1.52%

Effort curve

Score vs. reported cost. Lines connect complete sweeps only.

0 values
No cost-linked effort sweep for this version.
Definition and comparison boundarypublic methodology

Can the model solve difficult problems that require more than factual recall?

Multi-step reasoning on academic, mathematical, or abstract problems. Higher is better. The official ARC score ratio is converted to the catalog percent unit by multiplying by 100.

ARC-AGI version, semi-private split, exact source model label, reasoning setting, provider, release date, display status, and cost field are material. ARC-AGI-1, ARC-AGI-2, and ARC-AGI-3 must remain separate contexts.

Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID anthropic-claude-opus-5-high; exact model label Claude Opus 5 (High); model type CoT; model group anthropic-claude-opus-5; provider ID Anthropic; provider label Anthropic; release date 2026-07-24T00:00:00.000Z; raw score ratio 0.3016; catalog percent is raw score multiplied by 100; publisher raw cost 20657.37; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/anthropic-claude-opus-5. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID anthropic-opus-4-6-max-effort; exact model label Anthropic Opus 4.6 (Max); model type CoT; model group anthropic-opus-4-6-max-effort; provider ID Anthropic; provider label Anthropic; release date 2025-12-17T00:00:00.000Z; raw score ratio 0.0051; catalog percent is raw score multiplied by 100; publisher raw cost 8866.2; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-15. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID anthropic-opus-4-7-high; exact model label Opus 4.7 (High); model type CoT; model group anthropic-opus-4-7; provider ID Anthropic; provider label Anthropic; release date 2026-04-16T00:00:00.000Z; raw score ratio 0.0018; catalog percent is raw score multiplied by 100; publisher raw cost 10000; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-15. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID anthropic-opus-4-8-high; exact model label Claude Opus 4.8 (High); model type CoT; model group anthropic-opus-4-8; provider ID Anthropic; provider label Anthropic; release date 2026-06-01T00:00:00.000Z; raw score ratio 0.0152; catalog percent is raw score multiplied by 100; publisher raw cost 10000; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-15. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-1-pro-preview; exact model label Gemini 3.1 Pro (Preview); model type CoT; model group google-gemini-3-1-pro-preview; provider ID Google; provider label Google; release date 2026-03-05T00:00:00.000Z; raw score ratio 0.0042; catalog percent is raw score multiplied by 100; publisher raw cost 2213.53; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-15. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-8-flash-high; exact model label Gemini 3.8 Flash (High); model type CoT; model group google-gemini-3-8-flash; provider ID Google; provider label Google; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.10366279223923121; catalog percent is raw score multiplied by 100; publisher raw cost 4398.2756895; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/google-gemini-3-8-flash. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-8-flash-high-provider-adapter; exact model label Gemini 3.8 Flash - Provider Adapter (High); model type CoT; model group google-gemini-3-8-flash-provider-adapter; provider ID Google; provider label Google; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.35004243350752434; catalog percent is raw score multiplied by 100; publisher raw cost 4521.7983525; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/google-gemini-3-8-flash. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-8-flash-low; exact model label Gemini 3.8 Flash (Low); model type CoT; model group google-gemini-3-8-flash; provider ID Google; provider label Google; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.059895604936972165; catalog percent is raw score multiplied by 100; publisher raw cost 2669.05677675; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/google-gemini-3-8-flash. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-8-flash-low-provider-adapter; exact model label Gemini 3.8 Flash - Provider Adapter (Low); model type CoT; model group google-gemini-3-8-flash-provider-adapter; provider ID Google; provider label Google; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.15071935405559647; catalog percent is raw score multiplied by 100; publisher raw cost 3158.1084112500002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/google-gemini-3-8-flash. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-8-flash-medium; exact model label Gemini 3.8 Flash (Medium); model type CoT; model group google-gemini-3-8-flash; provider ID Google; provider label Google; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.039878093830617266; catalog percent is raw score multiplied by 100; publisher raw cost 2727.66378825; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/google-gemini-3-8-flash. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID google-gemini-3-8-flash-medium-provider-adapter; exact model label Gemini 3.8 Flash - Provider Adapter (Medium); model type CoT; model group google-gemini-3-8-flash-provider-adapter; provider ID Google; provider label Google; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.24202718056957026; catalog percent is raw score multiplied by 100; publisher raw cost 3826.20432375; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/google-gemini-3-8-flash. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-4-2026-03-05-high; exact model label GPT-5.4 (High); model type CoT; model group openai-gpt-5-4-2026-03-05-thinking; provider ID OpenAI; provider label OpenAI; release date 2026-03-05T00:00:00.000Z; raw score ratio 0.0021; catalog percent is raw score multiplied by 100; publisher raw cost 5187.41; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-15. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-5-2026-04-23-high; exact model label GPT-5.5 (High); model type CoT; model group openai-gpt-5-5-2026-04-23-thinking; provider ID OpenAI; provider label OpenAI; release date 2026-04-23T00:00:00.000Z; raw score ratio 0.0043; catalog percent is raw score multiplied by 100; publisher raw cost 10000; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-35. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-luna-high; exact model label GPT-5.6 Luna (High); model type CoT; model group openai-gpt-5-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.001; catalog percent is raw score multiplied by 100; publisher raw cost 2482.44; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-luna-low; exact model label GPT-5.6 Luna (Low); model type CoT; model group openai-gpt-5-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0017; catalog percent is raw score multiplied by 100; publisher raw cost 2307.49; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-luna-max; exact model label GPT-5.6 Luna (Max); model type CoT; model group openai-gpt-5-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0018; catalog percent is raw score multiplied by 100; publisher raw cost 3189.42; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-luna-medium; exact model label GPT-5.6 Luna (Medium); model type CoT; model group openai-gpt-5-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0017; catalog percent is raw score multiplied by 100; publisher raw cost 2379.13; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-luna-xhigh; exact model label GPT-5.6 Luna (XHigh); model type CoT; model group openai-gpt-5-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0002; catalog percent is raw score multiplied by 100; publisher raw cost 2801.59; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-sol-high; exact model label GPT-5.6 Sol (High); model type CoT; model group openai-gpt-5-6-sol; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0215; catalog percent is raw score multiplied by 100; publisher raw cost 15176.07; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-sol. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-sol-low; exact model label GPT-5.6 Sol (Low); model type CoT; model group openai-gpt-5-6-sol; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0033; catalog percent is raw score multiplied by 100; publisher raw cost 12805.6; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-sol. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-sol-max; exact model label GPT-5.6 Sol (Max); model type CoT; model group openai-gpt-5-6-sol; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0778; catalog percent is raw score multiplied by 100; publisher raw cost 25064.11; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-sol. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-sol-medium; exact model label GPT-5.6 Sol (Medium); model type CoT; model group openai-gpt-5-6-sol; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0107; catalog percent is raw score multiplied by 100; publisher raw cost 12971.17; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-sol. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-sol-xhigh; exact model label GPT-5.6 Sol (XHigh); model type CoT; model group openai-gpt-5-6-sol; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0699; catalog percent is raw score multiplied by 100; publisher raw cost 19216.38; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-sol. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-terra-high; exact model label GPT-5.6 Terra (High); model type CoT; model group openai-gpt-5-6-terra; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0049; catalog percent is raw score multiplied by 100; publisher raw cost 5881.39; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-terra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-terra-low; exact model label GPT-5.6 Terra (Low); model type CoT; model group openai-gpt-5-6-terra; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0001; catalog percent is raw score multiplied by 100; publisher raw cost 5068.48; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-terra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-terra-max; exact model label GPT-5.6 Terra (Max); model type CoT; model group openai-gpt-5-6-terra; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.008; catalog percent is raw score multiplied by 100; publisher raw cost 7918.31; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-terra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-terra-medium; exact model label GPT-5.6 Terra (Medium); model type CoT; model group openai-gpt-5-6-terra; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0008; catalog percent is raw score multiplied by 100; publisher raw cost 5895.31; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-terra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-5-6-terra-xhigh; exact model label GPT-5.6 Terra (XHigh); model type CoT; model group openai-gpt-5-6-terra; provider ID OpenAI; provider label OpenAI; release date 2026-07-09T00:00:00.000Z; raw score ratio 0.0065; catalog percent is raw score multiplied by 100; publisher raw cost 6804.04; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-5-6-terra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-high; exact model label GPT-6 Astra (High); model type CoT; model group openai-gpt-6-astra; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.5481905163612231; catalog percent is raw score multiplied by 100; publisher raw cost 40704.722830000006; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-high-provider-adapter; exact model label GPT-6 Astra - Provider Adapter (High); model type CoT; model group openai-gpt-6-astra-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.99945712321257; catalog percent is raw score multiplied by 100; publisher raw cost 18816.63138; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-low; exact model label GPT-6 Astra (Low); model type CoT; model group openai-gpt-6-astra; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.17452190401133805; catalog percent is raw score multiplied by 100; publisher raw cost 38166.47030000001; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-low-provider-adapter; exact model label GPT-6 Astra - Provider Adapter (Low); model type CoT; model group openai-gpt-6-astra-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.9802838165829648; catalog percent is raw score multiplied by 100; publisher raw cost 21297.922310000005; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-max; exact model label GPT-6 Astra (Max); model type CoT; model group openai-gpt-6-astra; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.6271280210060628; catalog percent is raw score multiplied by 100; publisher raw cost 26097.501720000007; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-max-provider-adapter; exact model label GPT-6 Astra - Provider Adapter (Max); model type CoT; model group openai-gpt-6-astra-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.9855074288592295; catalog percent is raw score multiplied by 100; publisher raw cost 17331.95327; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-medium; exact model label GPT-6 Astra (Medium); model type CoT; model group openai-gpt-6-astra; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.3858757449733784; catalog percent is raw score multiplied by 100; publisher raw cost 48090.34598999999; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-medium-provider-adapter; exact model label GPT-6 Astra - Provider Adapter (Medium); model type CoT; model group openai-gpt-6-astra-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.9843861309255745; catalog percent is raw score multiplied by 100; publisher raw cost 19284.776630000004; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-none; exact model label GPT-6 Astra (None); model type CoT; model group openai-gpt-6-astra; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.35177880317424; catalog percent is raw score multiplied by 100; publisher raw cost 49791.11317000001; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-none-provider-adapter; exact model label GPT-6 Astra - Provider Adapter (None); model type CoT; model group openai-gpt-6-astra-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.9672021831785531; catalog percent is raw score multiplied by 100; publisher raw cost 23456.663940000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-xhigh; exact model label GPT-6 Astra (XHigh); model type CoT; model group openai-gpt-6-astra; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.5934295409146055; catalog percent is raw score multiplied by 100; publisher raw cost 37317.38768; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-astra-xhigh-provider-adapter; exact model label GPT-6 Astra - Provider Adapter (XHigh); model type CoT; model group openai-gpt-6-astra-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-02T00:00:00.000Z; raw score ratio 0.9844155844155843; catalog percent is raw score multiplied by 100; publisher raw cost 18147.191119999996; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-astra. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-high; exact model label GPT-6 Luna (High); model type CoT; model group openai-gpt-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0018024216987988248; catalog percent is raw score multiplied by 100; publisher raw cost 255.97928910000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-high-provider-adapter; exact model label GPT-6 Luna - Provider Adapter (High); model type CoT; model group openai-gpt-6-luna-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0037847731587690087; catalog percent is raw score multiplied by 100; publisher raw cost 223.89528660000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-low; exact model label GPT-6 Luna (Low); model type CoT; model group openai-gpt-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0003291205821227175; catalog percent is raw score multiplied by 100; publisher raw cost 228.8330075; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-low-provider-adapter; exact model label GPT-6 Luna - Provider Adapter (Low); model type CoT; model group openai-gpt-6-luna-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.002374807559887363; catalog percent is raw score multiplied by 100; publisher raw cost 214.29109010000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-max; exact model label GPT-6 Luna (Max); model type CoT; model group openai-gpt-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0010360997760319559; catalog percent is raw score multiplied by 100; publisher raw cost 262.2377791; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-max-provider-adapter; exact model label GPT-6 Luna - Provider Adapter (Max); model type CoT; model group openai-gpt-6-luna-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0058683868915549685; catalog percent is raw score multiplied by 100; publisher raw cost 237.1428722; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-medium; exact model label GPT-6 Luna (Medium); model type CoT; model group openai-gpt-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0019415284639354709; catalog percent is raw score multiplied by 100; publisher raw cost 241.2797313; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-medium-provider-adapter; exact model label GPT-6 Luna - Provider Adapter (Medium); model type CoT; model group openai-gpt-6-luna-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.003161200923817726; catalog percent is raw score multiplied by 100; publisher raw cost 213.71995940000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-none; exact model label GPT-6 Luna (None); model type CoT; model group openai-gpt-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0002791945504283949; catalog percent is raw score multiplied by 100; publisher raw cost 229.77104200000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-none-provider-adapter; exact model label GPT-6 Luna - Provider Adapter (None); model type CoT; model group openai-gpt-6-luna-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.00046440444355949666; catalog percent is raw score multiplied by 100; publisher raw cost 205.03171590000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-xhigh; exact model label GPT-6 Luna (XHigh); model type CoT; model group openai-gpt-6-luna; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.0016413333960356602; catalog percent is raw score multiplied by 100; publisher raw cost 262.6844075; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID openai-gpt-6-luna-xhigh-provider-adapter; exact model label GPT-6 Luna - Provider Adapter (XHigh); model type CoT; model group openai-gpt-6-luna-provider-adapter; provider ID OpenAI; provider label OpenAI; release date 2026-09-22T00:00:00.000Z; raw score ratio 0.005072886313635385; catalog percent is raw score multiplied by 100; publisher raw cost 223.17200050000002; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/openai-gpt-6-luna. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID xai-grok-4-20-beta-0309-reasoning; exact model label Grok 4.20 (Beta Reasoning); model type CoT; model group xai-grok-4-20-beta-0309-reasoning; provider ID xAI; provider label xAI; release date 2026-03-05T00:00:00.000Z; raw score ratio 0.0009; catalog percent is raw score multiplied by 100; publisher raw cost 3775; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. labelOffsetX=-100, labelOffsetY=-15. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID xai-grok-4-5-high; exact model label Grok 4.5 (High); model type CoT; model group xai-grok-4-5; provider ID xAI; provider label xAI; release date 2026-07-16T00:00:00.000Z; raw score ratio 0.003; catalog percent is raw score multiplied by 100; publisher raw cost 6892.87; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/xai-grok-4-5. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID xai-grok-4-5-low; exact model label Grok 4.5 (Low); model type CoT; model group xai-grok-4-5; provider ID xAI; provider label xAI; release date 2026-07-16T00:00:00.000Z; raw score ratio 0.0026; catalog percent is raw score multiplied by 100; publisher raw cost 8718.48; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/xai-grok-4-5. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID xai-grok-4-5-medium; exact model label Grok 4.5 (Medium); model type CoT; model group xai-grok-4-5; provider ID xAI; provider label xAI; release date 2026-07-16T00:00:00.000Z; raw score ratio 0.0032; catalog percent is raw score multiplied by 100; publisher raw cost 8458.3; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/xai-grok-4-5. Source feed https://arcprize.org/media/data/leaderboard/v3.json. Official ARC-AGI 3 v3_Semi_Private displayed leaderboard row; exact source model ID xai-grok-4-6-xhigh; exact model label Grok 4.6 (XHigh); model type CoT; model group xai-grok-4-6; provider ID xAI; provider label xAI; release date 2026-08-11T00:00:00.000Z; raw score ratio 0.0211; catalog percent is raw score multiplied by 100; publisher raw cost 5612; the v3 feed does not identify this field as per-task, so no cost-per-task value is attached. results URL /results/xai-grok-4-6. Source feed https://arcprize.org/media/data/leaderboard/v3.json.

Method / source ↗