STEP3-VL-10B Technical Report

2026-01-14

AI Summary

This work addresses the challenge of reconciling efficiency and performance in multimodal models at the 10-billion-parameter scale. The authors propose a unified, fully unfrozen vision–language pretraining architecture, pretrained on 1.2 trillion multimodal tokens and further refined through over 1,000 rounds of reinforcement learning. A novel test-time parallel coordinated reasoning mechanism (PaCoRe) is introduced to enable scalable synergy between perception and reasoning. Built upon Qwen3-8B, the resulting open-source model achieves state-of-the-art results—matching or surpassing significantly larger models and top proprietary systems such as Gemini 2.5 Pro—on multiple benchmarks, including MMBench (92.2%), MMMU (80.11%), AIME2025 (94.43%), and MathVision (75.95%).

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We present STEP3-VL-10B, a lightweight open-source foundation model designed to redefine the trade-off between compact efficiency and frontier-level multimodal intelligence. STEP3-VL-10B is realized through two strategic shifts: first, a unified, fully unfrozen pre-training strategy on 1.2T multimodal tokens that integrates a language-aligned Perception Encoder with a Qwen3-8B decoder to establish intrinsic vision-language synergy; and second, a scaled post-training pipeline featuring over 1k iterations of reinforcement learning. Crucially, we implement Parallel Coordinated Reasoning (PaCoRe) to scale test-time compute, allocating resources to scalable perceptual reasoning that explores and synthesizes diverse visual hypotheses. Consequently, despite its compact 10B footprint, STEP3-VL-10B rivals or surpasses models 10$\times$-20$\times$ larger (e.g., GLM-4.6V-106B, Qwen3-VL-235B) and top-tier proprietary flagships like Gemini 2.5 Pro and Seed-1.5-VL. Delivering best-in-class performance, it records 92.2% on MMBench and 80.11% on MMMU, while excelling in complex reasoning with 94.43% on AIME2025 and 75.95% on MathVision. We release the full model suite to provide the community with a powerful, efficient, and reproducible baseline.

Problem

• multimodal intelligence
• compact model
• vision-language synergy
• complex reasoning
• efficient foundation model

Innovation

• Parallel Coordinated Reasoning
• unified unfrozen pre-training
• multimodal foundation model
• reinforcement learning post-training
• vision-language synergy