PorTEXTO - BLEU-1: leaderboard

Metric: Mean BLEU-1 (%) across the four subsets; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 18 models tracked.

Top models

#ModelScore
1Qwen3.6-27B [28]81.5
2Gemma-4-31B-it [8]79.5
3Qwen3.5-9B [27]78.8
4Ministral3-Instruct-8B [13]67.9
5Gemma-4-E4B-it [8]67.5
6LLaVA-OneVision-1.5-8B [1]63.1
7Nemotron-Nano-12B-VL-V2 [5]59.5
8InternVL3.5-38B [32]59
9Bee-RL-8B [37]58.8
10AMALIA-VL-DPO [6]55.5

Interactive version: theaggregate.ai/benchmark?slug=portexto-bleu-1 · How It Works · Data refreshed daily, snapshot 2026-09-29.