VTC-Bench (Code Tools) - Measurement: leaderboard
Metric: Pass rate (%) on measurement (105 questions on size, position and shape), answering VTC-Bench's 680 visual questions (538 multiple-choice, 142 open-ended) that call for chains of image operations with 32 OpenCV-based tools invoked by writing code (Qwen-Agent, or the Thyme framework for the open tool-use models), up to 20 rounds; rule matching plus a GPT-4o judge for open answers; proprietary models in high reasoning mode; higher is better. Source: arxiv.org. Saturation forecast: Around 2028. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (High) | 54.29 | #77 (Gemini 3 Pro) |
| 2 | GPT-5.2 (High) | 50.48 | #105 (GPT-5.2) |
| 3 | Gemini 3 Flash (High) | 50.48 | #93 (Gemini 3 Flash) |
| 4 | Qwen 3 VL 235B A22B Instruct | 40 | #264 |
| 5 | Qwen 3 VL 8B Instruct | 39.05 | #401 |
| 6 | O3 (High) | 38.1 | #121 (O3) |
| 7 | GPT-4o | 37.14 | #333 |
| 8 | Qwen 3 VL 235B A22B (Thinking) | 36.19 | #228 (Qwen 3 VL 235B A22B) |
| 9 | Qwen 3 VL 30B A3B (Thinking) | 36.19 | #338 (Qwen 3 VL 30B A3B) |
| 10 | Qwen 3 VL 32B (Thinking) | 35.24 | #287 (Qwen 3 VL 32B) |
| 11 | Qwen 3 VL 8B (Thinking) | 32.38 | |
| 12 | Qwen 3 VL 32B Instruct | 31.43 | #276 |
| 13 | Qwen 3 VL 30B A3B Instruct | 31.43 | #365 |
| 14 | O4 Mini (High) | 31.43 | #172 (O4 Mini) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=vtc-bench-code-tools-measurement · How It Works · Data refreshed daily, snapshot 2026-10-11.