GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
One-RL-to-See-Them-All
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
One-RL-to-See-Them-All
Model Tree
Qwen/Qwen2.5-VL-7B-Instruct
Input Modalities
Output Modalities
Supported Functionality
Orsta-7B is a cutting-edge vision-language model (VLM) designed to achieve superior performance across a wide spectrum of both visual reasoning and visual perception tasks. This model is a result of post-training with V-Triune, our novel unified reinforcement learning (RL) system.
The V-Triune system enables VLMs to be jointly optimized on diverse multimodal tasks within a single, cohesive training pipeline. Orsta-7B has been specifically trained using V-Triune on a carefully curated set of eight challenging visual tasks, fostering robust generalization and enhanced capabilities.
Orsta-7B's advanced abilities stem from its training with the V-Triune system. Key aspects of its training include:
Unified RL Framework (V-Triune): V-Triune is a Visual Triple-Unified Reinforcement Learning system featuring three core complementary components:
Diverse Joint Task Optimization: Orsta-7B was jointly optimized on the following eight visual tasks:
This comprehensive training allows Orsta-7B to develop a deeper understanding of visual content and its relation to textual prompts, excelling in tasks that require intricate reasoning and precise perception.
| Model | Knowledge | Mathematics | Perception | Coding | Info. Ex. | Planning | Science | Metrics | MEGA-BenchCore |
|---|---|---|---|---|---|---|---|---|---|
| QwenVL-2-7B | 39.96 | 25.95 | 39.99 | 31.49 | 40.29 | 16.64 | 28.59 | 43.61 | 34.47 |
Orsta-7B is developed by post-training the Qwen2.5-VL-7B-Instruct model using our V-Triune reinforcement learning system. Consequently, its core usage, particularly regarding input formatting and model interaction, largely follows the established patterns of the Qwen2.5-VL series.
For comprehensive details on the base model's capabilities, multi-turn dialogue format, image input encoding specifics, and other functionalities, we recommend referring to the official Qwen2.5-VL documentation.
If you use Orsta-7B or the V-Triune system in your research, please cite our work:
bibtex
@article{ma2025one,title={One RL to See Them All: Visual Triple Unified Reinforcement Learning},author={Ma, Yan and Du, Linge and Shen, Xuyang and Chen, Shaoxiang and Li, Pengfei and Ren, Qibing and Ma, Lizhuang and Dai, Yuchao and Liu, Pengfei and Yan, Junjie},journal={arXiv preprint arXiv:2505.18129},year={2025}}
| QwenVL-2.5-7B | 38.84 | 27.67 | 41.24 | 28.93 | 50.23 | 16.32 | 36.75 | 41.64 | 35.07 |
| InternVL-3-8B | 36.64 | 32.75 | 42.17 | 35.11 | 48.92 | 14.35 | 36.51 | 53.94 | 36.48 |
| Gemma3-12B | 41.11 | 29.10 | 37.38 | 30.27 | 46.56 | 16.10 | 36.83 | 50.40 | 35.04 |
| Kimi-VL-A3B | 37.63 | 27.07 | 39.50 | 22.30 | 40.99 | 22.17 | 33.94 | 46.65 | 34.40 |
| MM-Eureka-7B 💡 | 40.12 | 31.59 | 39.71 | 28.75 | 49.32 | 16.64 | 37.25 | 46.39 | 35.96 |
| VL-Rethinker-7B 💡 | 40.65 | 30.08 | 42.02 | 29.87 | 52.03 | 17.83 | 36.82 | 46.90 | 37.25 |
| Kimi-VL-A3B-Thinking 💡 | 33.45 | 17.76 | 28.11 | 14.69 | 41.14 | 12.64 | 28.60 | 43.97 | 27.08 |
| Orsta-7B (Ours) 💡 | 41.65 | 31.48 | 43.84 | 32.82 | 54.07 | 17.83 | 36.91 | 41.66 | 38.31 |
| - | - | - | - | - | - | - | - | - | - |
| Δ (Ours - Backbone) | +2.8 | +3.8 | +2.6 | +3.9 | +3.8 | +1.5 | +0.2 | +0.0 | +3.2 |