Ling-mini-base-2.0-20T
Ling-mini-base-2.0-20T is an open-weights language model from Ant Group, released September 10, 2025. It has 17B parameters (mixture-of-experts, 2.3B active). The context window is 4K tokens.
Today, we are excited to announce the open-sourcing of Ling 2.0 — a family of MoE-based large language models that combine SOTA performance with high efficiency. The first released version, Ling-mini-2.0, is compact yet powerful.
Check it against your own PC →
Can you run Ling-mini-base-2.0-20T locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q6_K, GPU + experts in RAM, roughly 33 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 5.2 GB | 10.1 GB | ~33 |
| 8 GB VRAM | Runs · GPU + CPU experts | Q6_K (≈ lossless) | 7.2 GB | 8.1 GB | ~42 |
| 12 GB VRAM | Runs on GPU | IQ4_XS (good) | 10.3 GB | – | ~150 |
| 16 GB VRAM | Runs on GPU | Q5_K_M (excellent) | 13.4 GB | – | ~160 |
| 24 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 15.3 GB | – | ~230 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 15.3 GB | – | ~370 |
Ready-made quantizations on Hugging Face: mradermacher/Ling-mini-base-2.0-20T-i1-GGUF · mradermacher/Ling-mini-base-2.0-20T-GGUF
Benchmark results
Epoch AI has not published results for Ling-mini-base-2.0-20T yet. New models are usually evaluated within days to weeks.
Links
Announcement · Hugging Face: inclusionAI/Ling-mini-base-2.0-20T
More from Ant Group
Ring-2.6-1T · Ling-2.6-flash · Ring-mini-linear-2.0 · Ring-flash-linear-2.0 · Ring-flash-2.0 · Ling-flash-base-2.0-20T · Agentar-Fin-R1 32B · Agentar-Fin-R1 8B