Skip to content
← Models
Open weightsHugging Face/10 trustVerified
Idefics2 8B logo

Idefics2 8B

Hugging Face's 8B open multimodal model with strong OCR and document skills.

Visit official page →

About

Idefics2-8B is an 8-billion-parameter open multimodal model from Hugging Face, released on May 3, 2024. Built on Mistral-7B-v0.1 with a SigLIP vision encoder, it accepts arbitrary sequences of images and text and produces text, with major gains over Idefics1 in OCR, document understanding, and visual reasoning.

Released under Apache 2.0 in base, instruction-tuned, and chatty checkpoints, it handles native-resolution images up to 980x980 without forced square resizing and was integrated into Transformers 4.40. Launch coverage compared it favorably with DeepSeek-VL and LLaVA-NeXT-Mistral-7B.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

May 3, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

visual-question-answeringocrdocument-understandingchart-analysismultimodal-chat

Trust score — 8/10

/10 trustVerified

Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 112; Trending score 82/100 from researcher assessment.

Sources