TIGER-Lab / text-generation updated 1 year ago

VLM2Vec-Full

Name: VLM2Vec-Full
Author: TIGER-Lab

This repo contains the code and data for VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks. In this paper, we aimed at building a unified multimodal embedding model for any tasks. Our model is based on converting an existing well-trained VLM (Phi-3.5-V) into an embedding model.

Params

4.2 B

Context

131,072

Downloads 30d

599 K

Likes

Commercial use: allowed apache-2.0 Not gated SAFETENSORS 1 languages View on Hugging Face ↗

Download history

daily snapshots · 10 days

612 K575 K

Jul 28Jul 31Aug 3Aug 6

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

File	Quant	Size	Est. VRAM	Verdict on RTX 4090 · 24 GB
model.safetensors	bf16	8.3 GB	10.2 GB	✅ Runs comfortably
model.safetensors (bf16, full)	bf16 + 128K ctx	8.3 GB	19.6 GB	✅ Runs comfortably

Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Run it

copy-paste, exact tags checked against the Hub

~ · curl · api/v1

$ curl -s https://aimodelscomparison.com/api/v1/models/vlm2vec-full
{
  "hf_id": "TIGER-Lab/VLM2Vec-Full",
  "params_b": 4.15,
  "context_length": 131072,
  "license": { "id": "apache-2.0", "commercial": "yes" },
  "downloads_30d": 599002,
  "vram_estimates": [
    { "quant": "bf16", "gb": 10.2 }
  ],
  "updated_at": "2026-07-28T18:04:39Z"
}

est. VRAM —on RTX 4090 · 24 GBJSON API →

Specifications

Architecture: Phi3VForCausalLM
Parameters: 4.2 B
Tensor type: BF16
Context length: 131,072
Vocabulary: 32,064
Layers / heads: 32 / 32
Licence: apache-2.0
First seen on the Hub: 2024-10-08
Base model: Phi-3.5-vision-instruct
Training datasets: TIGER-Lab/MMEB-train
Added to our catalog: 2026-07-28

Family

Base model and the most-downloaded derivatives in the catalog.

Base Phi-3.5-vision-instruct microsoft 1.3 M

Compare with

Qwen3-8B 16.3 M Qwen2.5-1.5B-Instruct 14.0 M Qwen2.5-7B-Instruct 12.1 M Llama-3.1-8B-Instruct 8.0 M Qwen2.5-3B-Instruct 6.0 M Qwen3-4B 4.6 M

Appears in

Best open coding models Commercial-use text models Models that fit in 12 GB