Bonsai 2 9B vs Qwen 3.5 9B
Side-by-side VRAM requirements, benchmark scores, and GPU compatibility for local AI inference.
Quick verdict
Bonsai 2 9B is more hardware-efficient: it needs 3.0 GB at its Ternary (PQ2_0) build vs 6.4 GB for Qwen 3.5 9B's Q4_K_M, fitting on 117 GPUs natively.
VRAM at each real build (8k context)
Bonsai 2 9B ships as a handful of fixed prebuilt files rather than a standard quantization ladder, so it has no real Q4_K_M/FP16/etc. equivalent to Qwen 3.5 9B's. Ranked smallest to largest instead.
| Rank | Bonsai 2 9B | GB | Qwen 3.5 9B | GB |
|---|---|---|---|---|
| 1 | Ternary (PTQ1_0) | 2.5 GB | Q2_K | 4.1 GB |
| 2 | Ternary (PQ2_0) | 3.0 GB | Q3_K_M | 5.2 GB |
| 3 | Ternary (Q2_0) | 3.2 GB | NVFP4 | 5.3 GB |
| 4 | — | — | Q4_K_M | 6.4 GB |
| 5 | — | — | Q5_K_M | 7.5 GB |
| 6 | — | — | Q6_K | 8.6 GB |
| 7 | — | — | Q8_0 | 11.0 GB |
| 8 | — | — | BF16 | 20.5 GB |
| 9 | — | — | FP16 | 20.5 GB |
| 10 | — | — | FP32 | 40.6 GB |
Each column is that model's own real builds, smallest first; rank pairs them by position, not by matching precision.
Model specifications
| Spec | Bonsai 2 9B | Qwen 3.5 9B |
|---|---|---|
| Org | PrismML | Alibaba |
| Parameters | 9B | 9B |
| Architecture | Dense | Dense |
| Context | 256k tokens | 256k tokens |
| Modalities | text, vision | text, vision |
| License | Apache 2.0 | Apache 2.0 |
| Commercial | Yes | Yes |
| Released | 2026-10-15 | 2026-02-24 |
| GPUs (native) | 117 / 121 | 116 / 121 |
Benchmark scores
| Benchmark | Bonsai 2 9B | Qwen 3.5 9B |
|---|---|---|
| GPQA Diamond | 69.9 | 81.7 |
| IFEval | 90.0 | 91.5 |
Green = higher score (better). N/A = not yet available. ~ = inherited from a base model, not independently reported for that release itself.
GPUs that run only Bonsai 2 9B(1)
GPUs that run only Qwen 3.5 9B(0)
Every GPU that runs Qwen 3.5 9B also runs Bonsai 2 9B.
GPUs that run both natively(116)
- NVIDIA RTX 509032 GB
- NVIDIA RTX 508016 GB
- NVIDIA RTX 5070 Ti16 GB
- NVIDIA RTX 507012 GB
- NVIDIA RTX 5060 Ti 16GB16 GB
- NVIDIA RTX 5060 Ti 8GB8 GB
- NVIDIA RTX 50608 GB
- NVIDIA RTX 50508 GB
- NVIDIA RTX 409024 GB
- NVIDIA RTX 408016 GB
- NVIDIA RTX 4070 Ti SUPER16 GB
- NVIDIA RTX 4070 Ti12 GB
- +104 more GPUs run both
Which should you use?
- • It's the newer release (2026-10-15 vs 2026-02-24); check the benchmark table above for what actually improved
- • You need chain-of-thought reasoning
Frequently asked questions
- Which is better, Bonsai 2 9B or Qwen 3.5 9B?
- Bonsai 2 9B is more hardware-efficient, needing 3.0 GB at its Ternary (PQ2_0) build vs 6.4 GB for Qwen 3.5 9B's Q4_K_M. Bonsai 2 9B runs on more GPUs natively (117 vs 116).
- How much VRAM does Bonsai 2 9B need vs Qwen 3.5 9B?
- At 8k context, Bonsai 2 9B needs approximately 3.0 GB of VRAM at its Ternary (PQ2_0) build, while Qwen 3.5 9B needs 6.4 GB at its Q4_K_M build. At the largest build each ships, Bonsai 2 9B requires 3.2 GB (Ternary (Q2_0)) vs 20.5 GB (FP16) for Qwen 3.5 9B.
- Can you run Bonsai 2 9B on the same GPUs as Qwen 3.5 9B?
- Yes, 116 GPUs can run both natively in VRAM, including NVIDIA RTX 5090, NVIDIA RTX 5080, NVIDIA RTX 5070 Ti. However, 1 GPU can run Bonsai 2 9B but not Qwen 3.5 9B, and no GPU can run Qwen 3.5 9B without also fitting Bonsai 2 9B.
- What is the difference between Bonsai 2 9B and Qwen 3.5 9B?
- Bonsai 2 9B has 9B parameters (dense) with a 256k context window. Qwen 3.5 9B has 9B parameters (dense) with a 256k context window.
- Which model fits in 24 GB of VRAM, Bonsai 2 9B or Qwen 3.5 9B?
- Both fit in 24 GB of VRAM at their respective recommended builds: Bonsai 2 9B (Ternary (PQ2_0)) needs 3.0 GB and Qwen 3.5 9B (Q4_K_M) needs 6.4 GB.