Idefics3
Open vision-language model that doubles as a VLM building tutorial.
About
Idefics3 is Hugging Face's 8-billion-parameter vision-language model, pairing SigLIP vision with a Llama-3-based text model and dropping the perceiver module used in Idefics2. The team released it alongside Docmatix, a document-understanding dataset they describe as 240 times larger than what came before, plus the full training recipe. Checkpoints are Apache-2.0, though the underlying Llama 3.1 base carries its own terms. It is a solid open baseline for document-heavy vision tasks, but it is a 2024 model and has been overtaken by newer open VLMs on several benchmarks.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Reported to significantly outperform Idefics2-8B across the paper's evaluation suite (authors, 2024).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: Released 2024-08-22; integrated into Hugging Face Transformers 2024-09-25; Checkpoints Apache-2.0 per the model card; Built on Llama 3.1 8B Instruct, whose own license terms also apply; Outperforms-Idefics2 claims are author-reported from the 2024 paper. License: Checkpoints Apache-2.0; underlying Llama 3.1 base carries Meta's license terms..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · Released 2024-08-22; integrated into Hugging Face Transformers 2024-09-25
- · Checkpoints Apache-2.0 per the model card
- · Built on Llama 3.1 8B Instruct, whose own license terms also apply
- · Outperforms-Idefics2 claims are author-reported from the 2024 paper