LLM Reference

Kosmos 2

Released
2023-03-15
Last refreshed
2026-05-19
Status
Researched 105d ago
DeprecatedOpen sourceCommercial use: permitted

Kosmos 2 is a legacy integration reference; keep it only while you identify a current replacement.

Use it for

  • Teams maintaining an existing integration
  • Workloads that can use a 2k context window
  • Buyers comparing 1 tracked provider route

Do not use it for

  • New production launches
  • Vision or document-understanding workloads
  • Strict JSON or tool-calling flows
Specifications
Family
Kosmos-2
Released
2023-03-15
Context
2k
Parameters
1.66B
Architecture
Decoder Only
Specialization
general
Openness
Open source
License
MITOSI-approvedCommercial use: permitted
Weights
Unknown
Code
Unknown
Training
Fine-tuned
Created by

Advancing the state-of-the-art in AI and computing.

Redmond, Washington, United States
Founded 1991
Website
Pricing
Output / 1M
-
Input / 1M
-

Cheapest of 1 route · NVIDIA NIM

About

Kosmos-2, developed by Microsoft Research, is an advanced multimodal large language model (MLLM) that enhances the capabilities of its predecessor, Kosmos-1. It features a Transformer-based architecture trained on the GrIT dataset of grounded image-text pairs, enabling it to understand and interact with both text and visual data. A key innovation is Kosmos-2's ability to ground language to the visual world, allowing for nuanced interaction with images by linking text to specific visual elements using location tokens.

Top use-case fit

No primary decision-task fit is mapped for this model yet.

Provider price ladder

Compare API pricing across 1 providers for input and output tokens, batch, and cached reads when available.

ProviderInput / 1MOutput / 1MRoute
NVIDIA NIM--
ProvisionedPartial

Available via routers & gateways(1)

Capabilities

No model capability flags are currently sourced.

Benchmark peer barsfor Coding

No task-mapped benchmark peers are available for this model yet.

Migration checks

No linked migration route is available for this model yet.