Reasoning
ARC-AGI Leaderboard
ARC-AGI tests fluid, abstract reasoning on small visual grid puzzles where each task follows a novel rule the model must infer from a few examples. It deliberately resists memorization and is one of the most-watched measures of general reasoning progress.
Source: epoch19 open models ranked+213 proprietaryData through Sep 2026
Open models ranked on ARC-AGI
# shows rank among open models / rank overall (including proprietary).
| # | Model | Score | VRAM @ Q4 |
|---|---|---|---|
| 1 / 36 | DeepSeek V4.1 Flash · 763.2B | 94.5% | ~458 GB |
| 2 / 39 | Kimi K3 · 2779.9B | 94.5% | ~1674 GB |
| 3 / 60 | GLM 5.3 Flash · 321.3B | 91.0% | ~195 GB |
| 4 / 63 | DeepSeek V4 Pro 0813 · 1650.5B | 90.5% | ~991 GB |
| 5 / 71 | DeepSeek V4 Flash 0731 · 304.2B | 89.0% | ~183 GB |
| 6 / 77 | Qwen3.8 27B · 27.8B | 87.5% | ~17 GB |
| 7 / 91 | Inkling Small · 266.0B | 84.0% | ~160 GB |
| 8 / 96 | Inkling · 952.4B | 79.5% | ~572 GB |
| 9 / 99 | GLM 5.2 · 753.3B | 77.0% | ~456 GB |
| 10 / 119 | Kimi K2.5 · 1026.9B | 65.3% | ~620 GB |
| 11 / 122 | MiniMax M2.5 · 228.7B | 63.7% | ~138 GB |
| 12 / 132 | DeepSeek V3.2 · 685.4B | 57.0% | ~415 GB |
| 13 / 146 | GLM 5 · 753.9B | 44.7% | ~457 GB |
| 14 / 194 | DeepSeek R1 0528 · 684.5B | 21.2% | ~415 GB |
| 15 / 205 | DeepSeek R1 · 684.5B | 15.8% | ~415 GB |
| 16 / 215 | Qwen3 235B A22B Instruct 2507 · 235.1B | 11.0% | ~142 GB |
| 17 / 225 | Magistral Small 2506 · 23.6B | 5.0% | ~15 GB |
| 18 / 227 | Llama 4 Maverick 17B 128E Instruct · 401.6B | 4.4% | ~265 GB |
| 19 / 231 | Llama 4 Scout 17B 16E Instruct · 108.6B | 0.5% | ~72 GB |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- Magistral Small 2506, 24B, score 5.0% — on the efficiency frontier (best score at its size or smaller).
- Qwen3.8 27B, 28B, score 87.5% — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4 Flash 0731, 304B, score 89.0% — on the efficiency frontier (best score at its size or smaller).
- GLM 5.3 Flash, 321B, score 91.0% — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4.1 Flash, 763B, score 94.5% — on the efficiency frontier (best score at its size or smaller).
ARC-AGI: frequently asked questions
- What is the best open LLM on ARC-AGI?
- DeepSeek V4.1 Flash is the top open model on ARC-AGI, scoring 94.5%. Among all models tested — including proprietary ones — it ranks #35. The top model overall is Claude Fable 5 (xhigh) (Anthropic) at 98.5%.
- What's the best ARC-AGI model you can run on a 24 GB GPU?
- Qwen3.8 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (~17 GB), scoring 87.5% on ARC-AGI.
- Can open models match proprietary models on ARC-AGI?
- Not quite on ARC-AGI: the strongest proprietary model (Claude Fable 5 (xhigh)) scores 98.5%, ahead of the best open model (DeepSeek V4.1 Flash) at 94.5% — but you can run the open one yourself.
Scores aggregated from epoch. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.