23.6B
- Released
- 1y ago
- Size
- 33.3Bparameters
Ai2 mixture-of-experts model whose experts are trained separately on private data and can be added or removed later.
How it compares
Frequently asked questions
Running locally
FlexOlmo 7x7B has 33.3B parameters and needs an estimated 23.7 GB of VRAM at Q4, 41.2 GB at Q8 and 77 GB at FP16. All figures are estimates from the parameter count, covering weights plus about 15% and 0.5 GB for a short context.
Yes, it fits on a 24 GB GPU at Q4 (estimated 23.7 GB). It does not fit on a 16 GB, 12 GB or 8 GB GPU at Q4, Q8 or FP16, based on estimates.
Yes, assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits the estimated 23.7 GB at Q4 is 32 GB. This is an estimate.
Privacy and data
Yes, FlexOlmo 7x7B has open weights, so you can self-host it for full control of your data.
Capabilities and specs
No, FlexOlmo 7x7B only accepts text input.
Quality and benchmarks
The closest alternatives, matched on model kind and release date with no comparable price, are Devstral Small by Mistral AI, Kimi K2 Instruct by Moonshot AI and Jamba Large by AI21 Labs.
Devstral SmallFlexOlmo 7x7B vs Devstral SmallKimi K2 InstructFlexOlmo 7x7B vs Kimi K2 InstructJamba LargeFlexOlmo 7x7B vs Jamba Large
Licensing and access
FlexOlmo 7x7B has open weights and is licensed under Apache 2.0.
Yes, FlexOlmo 7x7B is licensed under Apache 2.0, a permissive license that allows commercial use.
The weights are available on the Hugging Face repository allenai/FlexOlmo-7x7B-1T.
Basics
FlexOlmo 7x7B was made by Ai2 and released on July 9, 2025.
Yes, FlexOlmo 7x7B has open weights.
Generated from Vertical Tabs data, updated November 3, 2026.
Discussion
No comments yet
Be the first to share your thoughts.