MLX 8-bit conversion of IFM/K2-Horizon-7B, a 7B dense decoder-only model for reasoning, coding, long-context work, and tool use. The source checkpoint supports a native context length of 524,288 tokens (512K).
Benchmark results reported by IFM for the original IFM/K2-Horizon-7B checkpoint.
| Format | Quantization | Size |
|---|---|---|
| MLX safetensors (2 shards) | Affine 8-bit, group size 64 | 9.56 GB |
The included model.py loader preserves K2-Horizon's grouped RMSNorm. Use it with MLX-LM and --trust-remote-code. The model is text-only; no vision projector or MTP files are included.
pip install -U mlx-lm
mlx_lm.generate \
--model abenzerps/K2-Horizon-7B-MLX-8bit \
--trust-remote-code \
--prompt "Explain why reproducible builds matter." \
--max-tokens 256
2c9659amodel-00001-of-00002.safetensors 5.3 GBmodel-00002-of-00002.safetensors 4.2 GBIFM/K2-Horizon-7B