Downloads · 30 days
49
14% of all-time downloads
ZZRI/Feihua-n2-2.6B-preview-GGUF
Feihua-n2-2.6B-preview-GGUF is a machine learning model from ZZRI. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as other.
Downloads · 30 days
49
14% of all-time downloads
All-time downloads
340
Public
Repo size
6.1 GB
Likes
0
Public
Click a slice to open those files.
.gguf6.1 GB · 100%
From the Hugging Face model README
Think clearly. Speak nonsense. Stop on time — no RL required. / 想清楚,说废话,该停就停,不用 RL。
</div>GGUF quantizations of Feihua-n2-2.6B-preview, the capacity-thesis release: the same dual-track SFT recipe that needed 8 GRPO episodes to cure thinking termination at 1.7B just works at 2.7B — 12/12 on the hardest four open-ended prompts, IFBench 25.0 family record, zero information conserved as always. Full research card at the main repo. Runs on llama.cpp mainline (lfm2 architecture).
本仓收录 Feihua-n2-2.6B-preview 的 GGUF 量化:参数量论实锤版——1.7B 要 8 回合 GRPO 才治好的思维链收尾,2.7B 纯 SFT 单发达成,最难四题 12/12,IFBench 25.0 家族新纪录,零信息量守恒依旧。完整研究报告见主仓。主线 llama.cpp(lfm2 架构)直接运行。
License / 许可:衍生自 LiquidAI/LFM2.5-2.6B,LFM Open License v1.0(见 LICENSE,署名 Liquid AI)。
| File / 文件 | Size / 大小 | Note / 建议 |
|---|---|---|
| feihua-n2-2p6b-preview-Q8_0.gguf | 2.8GB | Near-lossless / 准无损档 |
| feihua-n2-2p6b-preview-Q4_K_M.gguf | 1.7GB | Safe pick / 稳妥档 |
| feihua-n2-2p6b-preview-IQ4_XS.gguf | 1.0GB | Sweet spot / 甜点档 |
llama-server -m feihua-n2-2p6b-preview-Q4_K_M.gguf --reasoning-format deepseek -fa on
# 推荐采样 / recommended: temp 0.6, top_p 0.95
听君一席话,如听一席话。