Model reference · open weights

Ovis2

Available as managed deployment LLMs ATH-MaaS Vision + text 2 variants 15k dl/mo

Ovis2 is an open-weight language model from ATH-MaaS. AxForge deploys and operates it for you on dedicated EU-owned hardware — with the licence handled where one is required.

Available as managed deployment — configured and operated for you on dedicated EU hardware, quoted per deployment.

What it is

Released byATH-MaaS
TypeLanguage models
TaskVision + text
Parameters (lead)1.3B
Runs withtransformers
Released2025-02-10
Popularity15k downloads / month
LicenceOpen weights

About

What Ovis2 is

Introduction

GitHub | Paper

We are pleased to announce the release of Ovis2, our latest advancement in multi-modal large language models (MLLMs). Ovis2 inherits the innovative architectural design of the Ovis series, aimed at structurally aligning visual and textual embeddings. As the successor to Ovis1.6, Ovis2 incorporates significant improvements in both dataset curation and training methodologies.

Read the full model card

Key Features:

  • Small Model Performance: Optimized training strategies enable small-scale models to achieve higher capability density, demonstrating cross-tier leading advantages.

  • Enhanced Reasoning Capabilities: Significantly strengthens Chain-of-Thought (CoT) reasoning abilities through the combination of instruction tuning and preference learning.

  • Video and Multi-Image Processing: Video and multi-image data are incorporated into training to enhance the ability to handle complex visual information across frames and images.

  • Multilingual Support and OCR: Enhances multilingual OCR beyond English and Chinese and improves structured data extraction from complex visual elements like tables and charts.

Model Zoo

Ovis MLLMsViTLLMModel WeightsDemo
Ovis2-1Baimv2-large-patch14-448Qwen2.5-0.5B-InstructHuggingfaceSpace
Ovis2-2Baimv2-large-patch14-448Qwen2.5-1.5B-InstructHuggingfaceSpace
Ovis2-4Baimv2-huge-patch14-448Qwen2.5-3B-InstructHuggingfaceSpace
Ovis2-8Baimv2-huge-patch14-448Qwen2.5-7B-InstructHuggingfaceSpace
Ovis2-16Baimv2-huge-patch14-448Qwen2.5-14B-InstructHuggingfaceSpace
Ovis2-34Baimv2-1B-patch14-448Qwen2.5-32B-InstructHuggingface-

Performance

We use VLMEvalKit, as employed in the OpenCompass multimodal and reasoning leaderboard, to evaluate Ovis2.

Image Benchmark

BenchmarkQwen2.5-VL-3BSAIL-VL-2BInternVL2.5-2B-MPOOvis1.6-3BInternVL2.5-1B-MPOOvis2-1BOvis2-2B
MMBench-V1.1test77.173.670.774.165.868.476.9
MMStar56.556.554.952.049.552.156.7
MMMUval51.444.144.646.740.336.145.6
MathVistatestmini60.162.853.458.947.759.464.1
HallusionBench48.745.940.743.834.845.250.2
AI2D81.477.475.177.868.576.482.7
OCRBench83.183.183.880.184.389.087.3
MMVet63.244.264.257.647.250.058.3
MMBenchtest78.67772.876.667.970.278.9
MMT-Benchval60.857.154.459.250.855.561.7
RealWorldQA66.56261.366.75763.966.0
BLINK48.446.443.843.84144.047.9
QBench74.472.869.875.863.371.376.2
ABench75.574.571.175.267.571.376.6
MTVQA24.920.222.621.121.723.725.6

Video Benchmark

BenchmarkQwen2.5-VL-3BInternVL2.5-2BInternVL2.5-1BOvis2-1BOvis2-2B
VideoMME(wo/w-subs)61.5/67.651.9 / 54.150.3 / 52.348.6/49.557.2/60.8
MVBench67.068.864.360.3264.9
MLVU(M-Avg/G-A

From the published model card. Full card on the HuggingFace links in the sidebar.

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys ovis2 for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (ovis2 below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/chat/completions \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"ovis2","messages":[{"role":"user","content":"Hello"}]}'

Create an account — your API key is available in the console. 3M free tokens every 30 days with every new account.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms