Vector Eval - ARC-C: leaderboard
Vector Institute Inspect Evals run of ARC Challenge, testing harder science exam questions that require reasoning beyond simple retrieval.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O1 | 97.87 | #183 |
| 2 | Claude 3.5 Sonnet | 96.93 | #350 |
| 3 | DeepSeek R1 | 96.67 | #251 |
| 4 | O3 Mini | 96.42 | #272 |
| 5 | GPT-4o | 96.33 | #343 |
| 6 | Gemini 1.5 Pro | 96.33 | #445 |
| 7 | Qwen3 Coder Next | 96.16 | #326 |
| 8 | Llama 3.1 70B Instruct | 94.45 | #561 |
| 9 | Mistral Large 2 (Jul) | 94.37 | #484 |
| 10 | Gemini 1.5 Flash | 93.09 | #618 |
| 11 | GPT-4o Mini | 92.49 | #599 |
| 12 | Command-R+ | 85.07 | #891 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=vector-eval-arc-c · How It Works · Data refreshed daily, snapshot 2026-10-20.