The Qwen3.5-27B is a dense model from the series with 27 billion parameters, utilizing 64 layers and a hidden representation size of 4096. Unlike the MoE models in the series, the 27B does not use expert routing, which ensures more predictable behavior and stability in tasks requiring sequential logical inference. It retains a hybrid attention mechanism, ensuring efficient processing of long sequences (native context window of 262K tokens).
Thanks to full parameter activation, the model demonstrates superior results in tasks requiring adherence to complex instructions. Its score on the IFEval benchmark (95.0) is the highest in the medium-sized lineup, confirming its excellent ability to precisely follow user instructions. In mathematical reasoning (e.g., HMMT Feb 25 – 92.0) and programming (SWE-bench Verified – 72.4, LiveCodeBench v6 – 80.7), it achieves top-tier results, outperforming MoE-based architectures. Its multimodal capabilities are also top-notch: it holds the best result in the family on the BabyVision test (44.6) and is among the best in MathVision (86.0) and video understanding (VideoMME – 87.0).
The uniqueness of Qwen3.5-27B lies in its reliability and predictability for engineering tasks. It is the ideal choice for fintech applications, legal analysis, document automation, and building complex customer support chatbots, where accuracy and stability of responses are more critical than marginal computational savings. It stands out from MoE models due to its determinism and ease of optimization for specific tasks.
| Model Name | Context | Type | GPU | Status | Link |
|---|
There are no public endpoints for this model yet.
Rent your own physically dedicated instance with hourly or long-term monthly billing.
We recommend deploying private instances in the following scenarios:
| Name | GPU | TPS | Max Concurrency | |||
|---|---|---|---|---|---|---|
262,144.0 tensor |
3 | $0.88 | 1.054 | Launch | ||
262,144.0 tensor |
2 | $0.93 | 1.209 | Launch | ||
262,144.0 tensor |
3 | $1.06 | 1.054 | Launch | ||
262,144.0 tensor |
2 | $1.23 | 1.209 | Launch | ||
262,144.0 tensor |
2 | $1.56 | 1.209 | Launch | ||
262,144.0 tensor |
2 | $1.92 | 1.209 | Launch | ||
262,144.0 tensor |
2 | $2.22 | 2.101 | Launch | ||
262,144.0 |
1 | $2.37 | 3.148 | Launch | ||
262,144.0 tensor |
2 | $2.93 | 2.101 | Launch | ||
262,144.0 |
1 | $3.83 | 3.148 | Launch | ||
262,144.0 |
1 | $4.11 | 3.928 | Launch | ||
262,144.0 |
1 | $4.74 | 6.549 | Launch | ||
| Name | GPU | TPS | Max Concurrency | |||
|---|---|---|---|---|---|---|
262,144.0 tensor |
4 | $0.96 | 1.168 | Launch | ||
262,144.0 tensor |
4 | $1.26 | 1.168 | Launch | ||
262,144.0 tensor |
3 | $1.34 | 1.769 | Launch | ||
262,144.0 tensor |
2 | $2.22 | 1.478 | Launch | ||
262,144.0 tensor |
3 | $2.29 | 1.769 | Launch | ||
262,144.0 tensor |
4 | $2.34 | 2.952 | Launch | ||
262,144.0 |
1 | $2.37 | 2.524 | Launch | ||
262,144.0 tensor |
3 | $2.83 | 1.769 | Launch | ||
262,144.0 tensor |
2 | $2.93 | 1.478 | Launch | ||
262,144.0 |
1 | $3.83 | 2.524 | Launch | ||
262,144.0 |
1 | $4.11 | 3.305 | Launch | ||
262,144.0 |
1 | $4.74 | 5.925 | Launch | ||
| Name | GPU | TPS | Max Concurrency | |||
|---|---|---|---|---|---|---|
262,144.0 tensor |
6 | $1.65 | 1.217 | Launch | ||
262,144.0 tensor |
4 | $1.75 | 1.527 | Launch | ||
262,144.0 tensor |
4 | $2.34 | 1.527 | Launch | ||
262,144.0 |
1 | $2.50 | 1.100 | Launch | ||
262,144.0 tensor |
4 | $2.97 | 1.527 | Launch | ||
262,144.0 tensor |
4 | $3.68 | 1.527 | Launch | ||
262,144.0 tensor |
3 | $3.89 | 1.682 | Launch | ||
262,144.0 |
1 | $3.95 | 1.100 | Launch | ||
262,144.0 |
1 | $4.11 | 1.880 | Launch | ||
262,144.0 tensor |
3 | $4.34 | 1.682 | Launch | ||
262,144.0 |
1 | $4.74 | 4.500 | Launch | ||
Contact our dedicated neural networks support team at nn@immers.cloud or send your request to the sales department at sale@immers.cloud.