Allen Institute for AI (AI2)

About Allen Institute for AI (AI2)

Ai2 is a research institute publishing models, datasets, research tools and scientific applications. Its model work includes Olmo language models, Tülu post-training releases and Molmo multimodal models. It also maintains projects that support finding and working with research literature.

Focus

  • Open model and data research
  • Language and multimodal models
  • Scientific information tools

Projects and interfaces

Scope and limitations

  • An institute's wider project list does not imply that every project has a Sigma5C observation.

Official references

Models

48 named catalog records. These are name and distribution records, not a count of distinct model architectures.

  • Molmo2 8B
  • olmo-3
  • olmo-3:32b
  • olmo-3:32b-think
  • olmo-3:32b-think-fp16
  • olmo-3:32b-think-q4_K_M
  • olmo-3:32b-think-q8_0
  • olmo-3:7b
  • olmo-3:7b-instruct
  • olmo-3:7b-instruct-fp16
  • olmo-3:7b-instruct-q4_K_M
  • olmo-3:7b-instruct-q8_0
  • olmo-3:7b-think
  • olmo-3:7b-think-fp16
  • olmo-3:7b-think-q4_K_M
  • olmo-3:7b-think-q8_0
  • olmo-3:latest
  • olmo-3.1
  • olmo-3.1:32b
  • olmo-3.1:32b-instruct
  • olmo-3.1:32b-instruct-fp16
  • olmo-3.1:32b-instruct-q4_K_M
  • olmo-3.1:32b-instruct-q8_0
  • olmo-3.1:32b-think
  • olmo-3.1:32b-think-fp16
  • olmo-3.1:32b-think-q4_K_M
  • olmo-3.1:32b-think-q8_0
  • olmo-3.1:latest
  • olmo2
  • olmo2:13b
  • olmo2:13b-1124-instruct-fp16
  • olmo2:13b-1124-instruct-q4_K_M
  • olmo2:13b-1124-instruct-q8_0
  • olmo2:7b
  • olmo2:7b-1124-instruct-fp16
  • olmo2:7b-1124-instruct-q4_K_M
  • olmo2:7b-1124-instruct-q8_0
  • olmo2:latest
  • tulu3
  • tulu3:70b
  • tulu3:70b-fp16
  • tulu3:70b-q4_K_M
  • tulu3:70b-q8_0
  • tulu3:8b
  • tulu3:8b-fp16
  • tulu3:8b-q4_K_M
  • tulu3:8b-q8_0
  • tulu3:latest

Model families

  • OLMo (catalog grouping)
  • olmo2

Observations

Browse the Observatory

Browse the Observatory

AI-generated & AI-scored. Every observation on this site is produced and judged by AI, and AI makes mistakes. Treat each result as a recorded observation under a declared protocol, as of its date. Method and marker legend