|v0.14.0rc1|Latestreleasecandidate|See [QuickStart](https://docs.vllm.ai/projects/ascend/en/latest/quick_start.html) and [Installation](https://docs.vllm.ai/projects/ascend/en/latest/installation.html) formoredetails|
|v0.16.0rc1|Latestreleasecandidate|See [QuickStart](https://docs.vllm.ai/projects/ascend/en/latest/quick_start.html) and [Installation](https://docs.vllm.ai/projects/ascend/en/latest/installation.html) formoredetails|
|v0.13.0|Lateststableversion|See [QuickStart](https://docs.vllm.ai/projects/ascend/en/v0.13.0/quick_start.html) and [Installation](https://docs.vllm.ai/projects/ascend/en/v0.13.0/installation.html) formoredetails|
## Contributing
@@ -86,7 +86,7 @@ Below are the maintained branches:
|v0.13.0|最新正式/稳定版本|[快速开始](https://docs.vllm.ai/projects/ascend/en/v0.13.0/quick_start.html) and [安装指南](https://docs.vllm.ai/projects/ascend/en/v0.13.0/installation.html)了解更多|
This is the first release candidate of v0.16.0 for vLLM Ascend. Please follow the [official doc](https://docs.vllm.ai/projects/ascend/en/latest) to get started.
### Highlights
- Qwen3-Omni quantization adaptation and optimization is now available. [#6828](https://github.com/vllm-project/vllm-ascend/pull/6828)
- GLM5-W8A8 quantization is now supported by parameterizing hardcoded MLA dimensions. [#6902](https://github.com/vllm-project/vllm-ascend/pull/6902)
### Features
- [Experimental] Support FabricMem Mode for ADXL/HIXL interconnect. [#6806](https://github.com/vllm-project/vllm-ascend/pull/6806)
- Qwen3-Next now supports FlashComm1. [#6830](https://github.com/vllm-project/vllm-ascend/pull/6830)
- NPUWorker Profiler now supports profile_prefix for better profiling experience. [#6968](https://github.com/vllm-project/vllm-ascend/pull/6968)
- EPLB profiling now displays expert hotness comparison and time required for eplb adjustment. [#6877](https://github.com/vllm-project/vllm-ascend/pull/6877) [#7001](https://github.com/vllm-project/vllm-ascend/pull/7001)]
- Xlite Qwen3 MoE now supports Data Parallel. [#6715](https://github.com/vllm-project/vllm-ascend/pull/6715)
- Mooncake Layerwise Connector now supports kv_pool. [#7032](https://github.com/vllm-project/vllm-ascend/pull/7032)
- Eagle3 now supports QuaRot quantization without embedding. [#7038](https://github.com/vllm-project/vllm-ascend/pull/7038)
### Hardware and Operator Support
- 310P now supports w8a8sc quantization method. [#7075](https://github.com/vllm-project/vllm-ascend/pull/7075)
- Added AscendC casual_conv1d_fn operator for Qwen3-Next. [#6661](https://github.com/vllm-project/vllm-ascend/pull/6661)
- Added GMM custom operator for MoE models. [#7010](https://github.com/vllm-project/vllm-ascend/pull/7010)
### Performance
- Faster convolution computation improves TTFT by 0.95% and throughput by 0.59% for Qwen3-VL models. [#7017](https://github.com/vllm-project/vllm-ascend/pull/7017)
- Implement global CPU slicing and improve IRQ binding for Ascend NPUs, ensuring non-overlapping CPU partitions and better resource management. [#6945](https://github.com/vllm-project/vllm-ascend/pull/6945)
- Optimize MTP execution by reordering state update operation. [#6844](https://github.com/vllm-project/vllm-ascend/pull/6844)
- Avoid CPU sync in mrope_positions copy by using full tensor copy. [#7014](https://github.com/vllm-project/vllm-ascend/pull/7014)
- Remove H2D synchronization for expert_map in MoE models. [#7000](https://github.com/vllm-project/vllm-ascend/pull/7000)
### Dependencies
- CANN is upgraded to 8.5.1, please remember to upgrade by hand if you're not using the official image. [#6897](https://github.com/vllm-project/vllm-ascend/pull/6897)
### Deprecation & Breaking Changes
-`enable_flash_comm_v1` config option has been renamed back to `enable_sp`. [#6883](https://github.com/vllm-project/vllm-ascend/pull/6883)
- The auto-detect quantization format from model files is reverted, in v0.16.0rc1, we still need to add `---quantization ascend` to serve a model quantinized by modelslim. It will be added back in the next version after the bug with the remote model id is fixed. [#6873](https://github.com/vllm-project/vllm-ascend/pull/6873)
### Documentation
- Added user/developer guide for CPU binding. [#7045](https://github.com/vllm-project/vllm-ascend/pull/7045)
- Added metrics usage documentation and example. [#6962](https://github.com/vllm-project/vllm-ascend/pull/6962)
- Added llms.txt for LLM discovery. [#6886](https://github.com/vllm-project/vllm-ascend/pull/6886)
- Currently, for DeepSeek v3.2, PCP & DCP do not yet work with FlashComm1 feature, which may cause serve errors or other unknown errors.
- In 4-node A3 PD disaggregation deployment with DeepSeek V3.2, the P-Node may hang when benchmarking in high concurrency scenario, e.g., 2K/2K tokens with 512 concurrent requests.
- MTP with large EP configurations may cause graph capture buffer overflow. This is a bug need to fix in vLLM, now there is a workaround to avoid it: explicitly set `--compilation-config '{"max_cudagraph_capture_size": N}'` where `N = max_concurrency × (1 + num_speculative_tokens)`.
## v0.15.0rc1 - 2026.02.27
This is the first release candidate of v0.15.0 for vLLM Ascend. Please follow the [official doc](https://docs.vllm.ai/projects/ascend/en/latest) to get started.
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Canlin Guo