Compare / Ternary Bonsai 2 27B vs Qwen3.5 27B
Ternary Bonsai 2 27B vs Qwen3.5 27B
Updated November 2026
Verdict
Ternary Bonsai 2 27B scores higher on the Small LM Index: 74.6 vs 70.6, although their likely ranges overlap, so the lead is not certain. On the 3 benchmarks both labs report, Ternary Bonsai 2 27B wins 3 and Qwen3.5 27B wins 0. Both need about 18.2 GB of memory at 4-bit quantization. Only Qwen3.5 27B has a hosted API on OpenRouter ($0.69 per million tokens); the other has to be self-hosted.
Pick Ternary Bonsai 2 27B if…
- you want the strongest model overall
- you mostly write code
- you need math and reasoning
Pick Qwen3.5 27B if…
- you want a hosted API rather than running it yourself
- you want the more widely used model (more quantized builds and community support)
Specifications
| Ternary Bonsai 2 27B | Qwen3.5 27B | |
|---|---|---|
| Small LM Index | 74.6 (likely 72–77) | 70.6 (likely 69–72) |
| Lab | PrismML · United States | Alibaba (Qwen) · China |
| Released | 2026-09-17 | 2026-02-24 |
| Parameters | 27.4B (dense) | 27B (dense) |
| Memory at 4-bit | ~18.2 GB | ~17.9 GB |
| Context window | 262K tokens | 262K tokens |
| Input | text, image | text, image, video |
| Reasoning | Always thinks | Hybrid (thinking on/off) |
| License | Apache-2.0 | Apache-2.0 |
| API price / 1M tokens | Self-host only | $0.69 ($0.25 in, $2.00 out) |
| Hugging Face downloads (30 days) | — | 1.8M |
Benchmarks
Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.
| Benchmark | Ternary Bonsai 2 27B | Qwen3.5 27B |
|---|---|---|
| MMLU-Pro Knowledge | ≈ 86 | 86.1 |
| GPQA Diamond Knowledge | ≈ 85 | 85.5 |
| Humanity's Last Exam Reasoning | ≈ 24 | 24.3 |
| AIME Math | 95.8 | 92.6 |
| LiveCodeBench Coding | 90.1 | 80.7 |
| SWE-bench Verified Coding | ≈ 72 | 75.0 |
| Terminal-Bench Agentic | — | 41.6 |
| IFBench Instruction following | 74.0 | — |
| MMMU-Pro Vision | 75.5 | 75.0 |
| MMLU Knowledge | 89.1 | — |
| IFEval Instruction following | 91.3 | — |
| BFCL Agentic | 74.9 | — |
| MATH-500 Math | 98.8 | — |
Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.
Questions
Which is better, Ternary Bonsai 2 27B or Qwen3.5 27B?
Ternary Bonsai 2 27B scores higher on the Small LM Index: 74.6 vs 70.6, although their likely ranges overlap, so the lead is not certain. On the 3 benchmarks both labs report, Ternary Bonsai 2 27B wins 3 and Qwen3.5 27B wins 0.
Which is cheaper to run?
Only Qwen3.5 27B has a hosted API on OpenRouter ($0.69 per million tokens); the other has to be self-hosted. Both need about 18.2 GB of memory at 4-bit quantization.
Which has the longer context window?
Both have a 262K-token context window.
Can I use Ternary Bonsai 2 27B and Qwen3.5 27B commercially?
Both are released under the Apache-2.0 license. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.