1bit.MONSTERDocs GitHub ↗

OLMo — AI2 Dense

AI2's fully-open OLMo-2. Distinct from most transformers: LayerNorm instead of RMSNorm, and learned positional embeddings (no RoPE).

Models

Model Params 1BP Size Backend(s) Perf
OLMo-2-7B 7B 3.9 GB GGML-Vulkan / ZINC / NPU / HIP
OLMo-2-13B 13B 7.6 GB GGML-Vulkan / ZINC / NPU / HIP

Notes

See also: full model support detail · benchmarks SSOT · all families