Kosmos-2

Microsoft · 2023 · Multimodal LLM · USA

A to Z

A · App: No
B · Built By: Microsoft
C · CEO & Politics: Satya Nadella — Centrist / moderate
D · Date: 2023
E · Expertise: Intermediate+
F · Free Tier: Yes
G · Generates: Text only
H · Deployment: Local (open weights)
I · Inputs: Text+Image+Audio
J · Jurisdiction: USA
K · Key Staffers: Mustafa Suleyman (CEO Microsoft AI), Kevin Scott (CTO)
L · Languages: English-primary
M · Model Type: Multimodal LLM
N · Safety Issues: None reported
O · Open Source: Yes — open weights
P · Pricing: Not sold (research model)
Q · AIDb Rating: [UNKNOWN]
R · What Makes It Special: Microsoft's Kosmos-2, a multimodal large language model that can perceive object descriptions such as bounding boxes and ground text in images, writing referring expressions as Markdown-style links to location tokens; trained on GrIT, a large set of grounded image-text pairs, it handles phrase grounding, referring-expression comprehension and generation, perception-language tasks and language tasks, with code and weights released.
S · Stock / Status: $516 (MSFT, Sep 25, 2026)
T · Training Cutoff: [UNKNOWN]
U · Users: [UNKNOWN]
V · Voice Chat: No
W · Weights: Downloadable
X · X Handle: @Microsoft
Y · Year Founded: [UNKNOWN]
Z · Bottom Line: Open-weights Multimodal LLM from Microsoft, released 2023.

Data completeness 85% · Last verified 2026-05-12 · Primary source ↗

“Kosmos-2.” AIDb — The AI Database. Verified 2026-05-12. https://theaidatabase.com/m/kosmos-2.html