Over the past five months since the release of Qwen2-VL, developers have built new models based on it, contributing valuable feedback. Now, Qwen2.5-VL introduces enhanced capabilities, including precise analysis of images, text, and charts, as well as object localization with structured JSON outputs. It understands long videos, identifies key events, and functions as an agent, interacting with tools on computers and phones. The model's architecture features dynamic video processing and an optimized ViT encoder for improved speed and accuracy.
Amazon Nova Micro is a text-only model optimized for cost and speed. With a context window of 128K tokens, it excels at tasks like text summarization, translation, interactive chat, and basic coding. Released as part of the Amazon Nova foundation models, it supports fine-tuning and distillation for customization on proprietary data.
| Qwen2.5-VL-32B | Nova Micro | |
|---|---|---|
Web Site
| ||
Provider
| ||
Chat
| ||
Release Date
| ||
Modalities
| text images video | text |
API Providers
| - | Amazon Bedrock |
Knowledge Cut-off Date
| Unknown | Purposefully not disclosed |
Open Source
| Yes (Source) | No |
Pricing Input
| $0 | $0.04 per million tokens |
Pricing Output
| $0 | $0.14 per million tokens |
MMLU
| 78.4% Source | 77.6% CoT Source |
MMLU-Pro
| 49.5% | - |
MMMU
| 70% | - |
HellaSwag
| Not available | - |
HumanEval
| Not available | 81.1% pass@1 Source |
MATH
| 82.2% | 69.3% CoT Source |
GPQA
| 46.0% Diamond | 40% Main Source |
IFEval
| Not available | 87.2% Source |
SimpleQA
| - | - |
AIME 2024 | - | - |
AIME 2025 | - | - |
Aider Polyglot
| - | - |
LiveCodeBench v5
| - | - |
Global MMLU (Lite)
| - | - |
MathVista
| - | - |
Mobile Application | - | - |
Compare AI. Test. Benchmarks. Mobile Apps Chatbots, Sketch
Copyright © 2025 All Right Reserved.