Qiniu
CoverageBenchmark
Benchmark Qwen3.5 397B A17B API latency, throughput, and cost efficiency. Compare response speed, token output, and pricing for scalable AI workloads.
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
Model details
Qwen3.5 397B A17B
Large-scale mixture-of-experts design with multimodal input and long-context support characterizes this model, which combines 397B total parameters with an active 17B expert configuration per token. The architecture is built to deliver production-grade inference, integrating multimodal capability, long-context handling, multi-token prediction speculative decoding, and W8A8 quantized deployment paths for efficient serving. This design choice lets the model balance broad knowledge capacity against the per-token compute needed for responsive responses, making it well suited for complex reasoning workloads across text and image inputs.
Practical deployment is a central focus, with validation covering single-node and multi-node online deployment, Prefill-Decode disaggregation, and functional, accuracy, and performance evaluation pipelines. Supported feature matrices include BF16 and W8A8 quantization, chunked prefill, automatic prefix caching, speculative decoding, asynchronous scheduling, tensor parallelism, and expert-level parallelism for distributed serving. The model is optimized for Ascend hardware starting from vllm-ascend v0.17.0rc1, with Ascend95DT support from v0.23.0rc1, making it a strong fit for teams building scalable, low-latency inference systems that require multimodal inputs, extended context windows, and structured reasoning outputs.
Qiniuqwen3.5-397b-a17b
Quick Info
Powered by- Provider
- Qiniu
- Model key
- qwen3.5-397b-a17b
- Release date
- Feb 22, 2026
- Last updated
- Feb 22, 2026
- Input modalities
- Output modalities
- Capabilities
Limits
- Output tokens
- 64,000 tokens
- Context window
- 256,000 tokens
Latest news about Qwen3.5 397B A17B
Qiniu
Coverage
Alibaba announced on February 16, 2026 the open-source release of Qwen3.5, with Qwen3.5-397B-A17B (also called "Qwen3.5-Plus") as the first model in the series. The page explicitly names the 397B-A17B variant and describes it as a natively multimodal foundation model targeting reasoning, coding, agent capabilities, and According to the Alibaba announcement, Qwen3.5-397B-A17B is trained on trillions of vision-language tokens across multilingual text, images, videos, STEM, and reasoning data, processes text, image, and video, generates text, and supports 201 languages and dialects (up from 119 in Qwen3). The model combines a linear att
Qiniu
CoverageBenchmark
InferenceX's technical analysis explicitly names Qwen3.5-397B-A17B as the first open-weights model in Alibaba's Qwen3.5 series, a native vision-language foundation model with 397B total parameters and 17B activated per forward pass, announced February 2026 with weights published on Hugging Face under Apache 2.0. The mo The page details that the checkpoint unifies thinking and non-thinking behavior in one model, operating in thinking mode by default with ... blocks, and drops Qwen3's /think and /nothink soft switches in favor of API parameters. Independent evaluation placed it at 45 on the Artificial Analysis Intelligence Index (3rd a
Qiniu
CoverageBenchmark
OpenRouter's Qwen3.5 397B A17B listing records a February 16, 2026 release with a 262K-token context window and a base price of $0.39 per million input tokens and $2.34 per million output tokens, with a weighted-average effective price of $0.442 input / $3.445 output per million tokens across providers. Alibaba Cloud I The model is described as a hybrid linear-attention plus sparse mixture-of-experts vision-language architecture targeting language understanding, logical reasoning, code generation, agent tasks, image and video understanding, and GUI interactions. Notably absent from OpenRouter's provider roster is Qiniu, confirming th