Sakura-Micro-Bonsai-2-GSQ-RCO

Sakura Micro Bonsai 2 GSQ-RCO

<img src="assets/sakura-logo.svg" alt="Sakura Micro Bonsai logo" width="280" />

Sakura Micro Bonsai 2 GSQ-RCO is an experimental ultra-low-bit derivative of Prism ML's Ternary Bonsai 2 27B, itself derived from Qwen3.8-27B. This release applies GSQ/RCO-based mixed-codec allocation and additional low-bit compression.

Provenance / Base model

Sakura Micro Bonsai 2 GSQ-RCO is derived from Prism ML's Ternary Bonsai 2 27B, which identifies Qwen3.8-27B as its base model.

Release artifact

Sakura-Micro-Bonsai-2-GSQ-RCO-TQ1_0-5.3GiB.gguf

The file name carries TQ1_0 because that ternary type holds most of the weights (394 of 851 tensors, 5.14 of 5.30 GiB); the remaining tensors use other, smaller types. The name was changed from Sakura-Micro-Bonsai-2-GSQ-RCO-5.3GiB.gguf on 2026-10-09 so that the Hub can show the quantization type; the file content is unchanged (same SHA-256).

• Size: 5,686,593,888 Bytes (5.296 GiB)
• SHA-256: 5054d9b3b3b95145ebf1def33803fd93a1962dda37ca50bd53dbcbbbe4655eb3
• The bundled SHA256SUMS file can be used to verify the downloaded GGUF.

Benchmark comparison

Sakura Micro Bonsai 2 GSQ-RCO is 4.37% smaller than the 5,946,648,928-Byte Ternary Bonsai 2 PTQ1 reference. These are measurements on small benchmark subsets, not a claim of general quality superiority over the base model.

Compression approach

The project combines several ideas rather than applying one uniform quantizer to every tensor. It uses GSQ/RCO-based mixed-codec allocation and additional low-bit compression to target a smaller release artifact while retaining measured functionality.

This card intentionally does not disclose per-tensor selections, allocation details, internal optimization logic, or reproduction procedures.

Notes

• Experimental low-bit release.
• Use the runtime documented by the base model.

License

Apache-2.0. The license text is included in LICENSE.

---

<details><summary>中文说明 · 樱花 (Simplified Chinese)</summary>

English above. 本节为上文的中文翻译(Sakura = 樱花 yīnghuā);完整的独立中文版见 README_zh.md。

Sakura Micro Bonsai 2 GSQ-RCO

<img src="assets/sakura-logo.svg" alt="Sakura Micro Bonsai logo" width="280" />

Sakura Micro Bonsai 2 GSQ-RCO 是 Prism ML 的 Ternary Bonsai 2 27B 的一个实验性超低比特衍生版本,而后者本身派生自 Qwen3.8-27B。本次发布采用基于 GSQ/RCO 的混合编码分配以及额外的低比特压缩。

来源 / 基础模型

Sakura Micro Bonsai 2 GSQ-RCO 派生自 Prism ML 的 Ternary Bonsai 2 27B,后者把 Qwen3.8-27B 标明为其基础模型。

发布文件

Sakura-Micro-Bonsai-2-GSQ-RCO-TQ1_0-5.3GiB.gguf

文件名带有 TQ1_0,因为这种三值类型承载了大部分权重(851 个张量中的 394 个,5.30 GiB 中的 5.14 GiB);其余张量使用其他较小的类型。文件名已于 2026-10-09 由 Sakura-Micro-Bonsai-2-GSQ-RCO-5.3GiB.gguf 改为此名,以便 Hub 显示量化类型;文件内容不变(SHA-256 相同)。

• 大小:5,686,593,888 字节(5.296 GiB)
• SHA-256:5054d9b3b3b95145ebf1def33803fd93a1962dda37ca50bd53dbcbbbe4655eb3
• 可使用随附的 SHA256SUMS 文件校验下载的 GGUF。

基准对比

Sakura Micro Bonsai 2 GSQ-RCO 比 5,946,648,928 字节的 Ternary Bonsai 2 PTQ1 参照小 4.37%。这些是在小规模基准子集上的测量,不是关于整体质量优于基础模型的声明。

压缩方法

该项目结合了多种思路,而不是对每个张量应用同一种统一的量化器。它使用基于 GSQ/RCO 的混合编码分配以及额外的低比特压缩,目标是在保持已测得功能的同时得到更小的发布文件。

本卡片有意不披露逐张量的选择、分配细节、内部优化逻辑或复现步骤。

说明

• 实验性的低比特发布。
• 请使用基础模型文档中说明的运行时。

许可证

Apache-2.0。许可证文本包含在 LICENSE 中。

</details>