These are NOT compatible with llama.cpp and its frontends and derivatives (LMStudio, Kobold, etc...). Use with ik_llama.cpp only.

Make sure to pass the -ger option to llama-server or llama-cli to enable grouped experts.

GGUF quants of Ring-mini-2.0

Using ik_llama.cpp (commit dbfd151594d723b8bed54326cb521d1bccfad802)

The importance matrix was generated with eaddario/imatrix-calibration combined_all_medium dataset and grouped experts option enabled.

All quants were generated/calibrated with the imatrix, including the K quants.

Compressed from BF16.

Ring-mini-2.0

🤗 Hugging Face | 🤖 ModelScope | 🐙 Experience Now

Today, we officially release Ring-mini-2.0 — a high-performance inference-oriented MoE model deeply optimized based on the Ling 2.0 architecture. With only 16B total parameters and 1.4B activated parameters, it achieves comprehensive reasoning capabilities comparable to dense models below the 10B scale. It excels particularly in logical reasoning, code generation, and mathematical tasks, while supporting 128K long-context processing and 300+ tokens/s high-speed generation.

Enhanced Reasoning: Joint Training with SFT + RLVR + RLHF

Built upon Ling-mini-2.0-base, Ring-mini-2.0 undergoes further training with Long-CoT SFT, more stable and continuous RLVR, and RLHF joint optimization, significantly improving the stability and generalization of complex reasoning. On multiple challenging benchmarks (LiveCodeBench, AIME 2025, GPQA, ARC-AGI-v1, etc.), it outperforms dense models below 10B and even rivals larger MoE models (e.g., gpt-oss-20B-medium) with comparable output lengths, particularly excelling in logical reasoning.

High Sparsity, High-Speed Generation

Inheriting the efficient MoE design of the Ling 2.0 series, Ring-mini-2.0 activates only 1.4B parameters and achieves performance equivalent to 7–8B dense models through architectural optimizations such as 1/32 expert activation ratio and MTP layers. Thanks to its low activation and high sparsity design, Ring-mini-2.0 delivers a throughput of 300+ tokens/s when deployed on H20. With Expert Dual Streaming inference optimization, this can be further boosted to 500+ tokens/s, significantly reducing inference costs for high-concurrency scenarios involving thinking models. Additionally, with YaRN extrapolation, it supports 128K long-context processing, achieving a relative speedup of up to 7x in long-output scenarios.