Ternary Bonsai 27B — 显存、速度与本地部署
作者： Jakub Rusinowski · 最后更新：
The laptop build and recommended pick when you have the memory. Ternary {−1, 0, +1} weights at ~1.71 bits/weight land at 5.9 GB while keeping more than 95% of the FP16 baseline. PrismML reports MATH-500 99.20, AIME 2025 90.84, HumanEval+ 93.90 and LiveCodeBench 82.8, with tool-use and agentic ability largely intact (~74). Multimodal, 262K context, Apache 2.0. Ships as GGUF and MLX; also hosted on Together AI. Specs from launch coverage — verify on the Hugging Face model card.
Ternary Bonsai 27B 在 Ternary (1.58-bit, ~1.71 bpw) 下约需 7 GB 显存——量化权重加框架开销，不含 KV 缓存。在 Apple Silicon 上，这部分来自统一内存。
按量化级别的显存与速度
计算基准：NVIDIA RTX 4090 (24 GB)。仅含权重与开销：该模型架构未公开，因此未计入 KV 缓存。
| 量化 | 显存 | 显存 | 速度（估算） | 适配 |
|---|---|---|---|---|
| Ternary (1.58-bit, ~1.71 bpw) 1.71 bpw | 6.6 GB | ~91 tok/s | 可运行 |
黑色标记 = NVIDIA RTX 4090 (24 GB) 上的可用显存。 估算来自内存带宽屋顶线模型，详见 方法说明页. Ternary Bonsai 27B 显存计算器 →
运行 Ternary Bonsai 27B
目录中能运行 Ternary Bonsai 27B 的最便宜 GPU 是 Intel Arc B570 (10 GB).
或在 Vast.ai 比较，低至 $0.35/小时 (typical low · varies)
作为亚马逊联盟成员，我们从符合条件的购买中获得收入。云 GPU 链接为推荐链接——我们可能获得佣金，您无需额外付费。
如何运行 Ternary Bonsai 27B
安装 Ollama，然后运行：
ollama run bonsai-27b规格
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
Preview — The model is released, but these specs are thin or rest on a single source. Individual fields may be wrong.
- 参数量
- 27 Billion
- 上下文窗口
- 262,144
- 架构
- Qwen3.6-27B (multimodal), ternary weights
- 提供商
- PrismML
- 许可证
- Apache 2.0
- Practical deployment
- Laptop
- 规格量化
- Ternary (1.58-bit, ~1.71 bpw)
- 系统内存
- 16 GB
- 记录更新于
- 2026-07-15
Commercial use permitted. No usage restrictions beyond attribution.
质量与使用场景
评分由模型作者或独立评测方发布——衡量质量而非吞吐量，并非我们实测。
| 基准 | 分数 | 来源 |
|---|---|---|
| MATH-500 | 99.2 % | 厂商声称 · PrismML (launch) |
| AIME 2025 | 90.84 % | 厂商声称 · PrismML (launch) |
| HumanEval+ | 93.9 % | 厂商声称 · PrismML (launch) |
| LiveCodeBench | 82.8 % | 厂商声称 · PrismML (launch) |
| Tool-use (agentic) | 74.01 % | 厂商声称 · PrismML (launch) |
| Vision | 65.19 % | 厂商声称 · PrismML (launch) |
我的 GPU 能运行 Ternary Bonsai 27B 吗？
- Ternary Bonsai 27B 在 AMD Radeon RX 9060 XT 8GB 上
- Ternary Bonsai 27B 在 Intel Arc B570 上
- Ternary Bonsai 27B 在 Intel Arc B580 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 3060 (12GB) 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 3070 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 3070 Ti 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 3080 (10GB) 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 4060 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 4070 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 4070 Super 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 4070 Ti 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 5060 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 5060 Ti 8GB 上
- Ternary Bonsai 27B 在 NVIDIA GeForce RTX 5070 上
Bonsai 27B 的其他尺寸
Ternary Bonsai 27B — 常见问题
How much VRAM does Ternary Bonsai 27B need?
About 7 GB at Ternary (1.58-bit, ~1.71 bpw) — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Ternary Bonsai 27B run on an RTX 4090 (24 GB)?
Yes. Ternary Bonsai 27B needs about 7 GB at Ternary (1.58-bit, ~1.71 bpw), inside a 24 GB card, at an estimated 91 tokens/sec.
How do I run Ternary Bonsai 27B locally?
Install Ollama and run `ollama run bonsai-27b`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.
What other sizes does Bonsai 27B come in?
1-bit Bonsai 27B (5 GB), Ternary Bonsai 27B (7 GB). Every size shares the family's training and licence; the larger ones score higher and need proportionally more memory.