LLMDB
Contribute
Shanghai AI Laboratory logo

InternVL 3

Shanghai AI LaboratoryOpen WeightsPending Human Review

InternVL 3 is Shanghai AI Laboratory’s open-weight vision-language model generation. A native multimodal pretraining generation with variable visual position encoding, stronger perception and reasoning, and multimodal test-time scaling. It supports language responses to images and video frames, document analysis, and visual questions. This family page groups ordinary size and packaging variants. Context and licensing depend on the selected checkpoint; large models can inherit additional language-backbone license conditions. Published evaluation results use specific visual preprocessing and inference settings, so they do not guarantee application performance.

2025-04-11
1B–78B
Vision Transformer, MLP projector, and decoder language model
MIT; backbone-specific licenses apply
Compare with other models

Specifications

Parameters
1B–78B
Architecture
Vision Transformer, MLP projector, and decoder language model
License
MIT; backbone-specific licenses apply
Type
multimodal
Modalities
textimagevideo

Benchmark Scores

Advanced Specifications

Model Family
InternVL
API Access
Not Available
Chat Interface
Not Available
Variants
InternVL3-78B

Capabilities & Limitations

Capabilities
visual question answeringdocument understandingvideo understandingreasoning
Known Limitations
Family members have different deployment and license conditionsHigh-resolution inputs consume contextVisual answers may hallucinate details
Notable Use Cases
document assistantsvideo analysismultimodal research

Resources

Related Models

Atria Dawn Preview

Shanghai AI Laboratory

Atria Dawn Preview

Atria Dawn Preview is Shanghai AI Laboratory’s agentic model for turning research and engineering questions into executable results. Built on GLM-5.2’s 744B mixture-of-experts foundation, it is post-trained for discovery, code creation, office deliverables, and cybersecurity tasks in authorized environments. The released text-only checkpoint supports a documented 256K inference context, downloadable MIT-licensed weights, and hosted API access. It combines tool execution with environmental feedback for experiment design, failure recovery, and verification. Configuration maximums above the documented serving limit are not treated as evidence of a larger supported context.

Typetext
Parameters744B
2026-09-14
Open Weights
DetailsCompare

Intern S2 397B

Shanghai AI Laboratory

Intern S2 397B

Intern S2 is the stable scientific multimodal follow-up to the 397B preview checkpoint. It strengthens extended agent execution and research workflows, with deep-thinking and ordinary-response modes. Public API documentation identifies the stable model separately and schedules preview decommissioning for October 31. The release includes downloadable Apache-licensed weights and chat access. The official laboratory announcement identifies the public release date as 2026-09-13.

Typemultimodal
Parameters397B
2026-09-13
Open Weights
DetailsCompare

Intern Lumina U2

Shanghai AI Laboratory

Intern Lumina U2

Intern Lumina U2 is a unified language-capable multimodal diffusion model covering text questions, visual understanding, image generation, and editing. Its sparse 16B backbone activates 1B parameters and uses an eight-codebook discrete visual tokenizer. The Ascend-trained weights are publicly available; the NVIDIA-trained checkpoint, training code, and report remain future work. Code publication in August preceded the September weight release. Primary catalogs verify the release month; the exact first-public day remains unverified and is not inferred from private repository or API creation metadata.

Typemultimodal
Parameters16B total, 1B active
2026-09
Open Weights
DetailsCompare