Qwen 3.5 27B
Qwen 3.5 27B is Alibaba’s dense 27-billion-parameter variant of the Qwen3.5 family (February 2026), an open Apache-2.0 release with native multimodality for text, image, and video. The hybrid architecture of Gated-DeltaNet and Gated-Attention blocks delivers 262,144 tokens of context (extensible to approximately one million via YaRN), up to 65,536 tokens of output, and strong coding performance (SWE-bench Verified 72.4) — the only dense variant in the Qwen3.5 family.
- Open Weights
- Workstation
- vLLM
- Text
- Vision
- Video
- Instruction-Tuned
- Long Context
- Unusable
Sovereign Risk: MEDIUM The model was developed by Alibaba Cloud in China. Although the weights are freely available under the Apache 2.0 license, the developer is subject to Chinese jurisdiction. This may raise considerations regarding data privacy and security for users in other legal jurisdictions. The risk is rated ‘medium’, as the open weights enable local use without transmitting data to the manufacturer.
Key metrics
Score · Latency · Cost · Quality
- Total Score Silver
- 71.01
- Routine
- 44.19
- Reasoning
- 26.82
- LLM Judge Avg
- 3.56 / 5
- 100 Coverage
- Avg Task Duration
- 190.65s
- Unusable
- Token Rate
- 17.78tok/s
- Output Rate
- P95 Latency
- 331.9s
- Top 5 %
- Total Tokens
- 182700
- Output Volume
- Cost per 1K
- $0
- USD / 1K Requests
- Benchmark Cost
- $0
- Total · 182700 tok
Benchmark modules
10 modules · weighted · vs. model median & top performer
Token efficiency & latency
Consumption per module vs. model median
Token consumption per module
Performance profile
Key metrics – legend
Score
- Total Score
- Overall score derived from Routine and Reasoning scores. Basis for the fleet rank and tier assignment (Gold / Silver / Bronze).
- Routine
- Score for standard tasks: code generation, text transformation, structured outputs. Weighted by task complexity.
- Reasoning
- Score for analytical and logical tasks: multi-step problem solving, mathematical reasoning, structured argumentation.
- Rank
- Position in the fleet comparison, sorted by Total Score. Refers to the snapshot at the time of testing.
LLM Judge
- LLM Judge Ø
- Average quality rating by an independent judge model on a scale of 1–5.
- Coverage
- Share of tasks for which a judge rating is available. 100% means all responses were evaluated.
Latency & speed
- Avg. Task Duration
- Average time per task from the first to the last token of the response, measured across all benchmark tasks.
- Token Rate
- Output speed in tokens per second. Relevant for interactive applications and real-time streaming.
- P95 Latency
- 95th percentile of response times — shows outliers. Lower values mean fewer long wait times in edge cases.
Cost
- Cost per 1K
- API price per 1,000 output tokens in US dollars. Basis for cost calculation in production use.
- Benchmark Cost
- Total cost for the complete benchmark run of this model.