Ternary Bonsai 2 27B
Ternary Bonsai 2 27B is available now for rag, agents, and long context with open-source and 256K context; evaluate it while provider pricing coverage matures.
- Family
- Bonsai
- Released
- 2026-09-17
- Context
- 256K
- Parameters
- 27.4B
- Specialization
- on-device
- Openness
- Open source
- License
- Apache 2.0OSI-approvedCommercial use: permitted
- Weights
- Available
- Code
- Unknown
- Training
- Fine-tuned
No tracked provider token pricing is available yet.
About
Ternary Bonsai 2 27B (also branded Bonsai 2 27B) is PrismML's flagship ternary open-weight multimodal model, announced 2026-09-17. Based on Qwen3.8 27B. PrismML reports ~9× smaller on-disk size than the full-precision counterpart and 98.2% of aggregate benchmark performance (overall 83.9 vs Qwen3.8 27B 85.4 on their thinking-mode suite). Language footprint ~5.9 GB (GGUF PTQ1_0 ~5.93–5.95 GB; PQ2_0 ~7.2 GB; MLX companion larger). Gains called out vs prior Bonsai 27B: agentic coding, multimodal reasoning, long-horizon tool use. Specs (docs/HF card): 27.36B params; 262K context; text+image in / text out; optional vision mmproj; Apache 2.0; available on Hugging Face (prism-ml/Ternary-Bonsai-2-27B-gguf + mlx-2bit).
Providers
No active provider routes are tracked for this model yet.
Capabilities
Benchmark peer barsfor RAG
No task-mapped benchmark peers are available for this model yet.
Migration checks
No linked migration route is available for this model yet.
API versions
ternary-bonsai-2-27bbonsai-2-27bNo tracked provider token pricing is available yet.