# Which ATTENTION BACKEND for gpt-oss in version 0.11.0?

**URL:** <https://discuss.vllm.ai/t/which-attention-backend-for-gpt-oss-in-version-0-11-0/1691>\
**Category:** Model Support\
**Created:** [October 4, 2025, 1:21am UTC](https://discuss.vllm.ai/t/which-attention-backend-for-gpt-oss-in-version-0-11-0/1691 "2025-10-04T01:21:28Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![thomas](https://avatars.discourse-cdn.com/v4/letter/t/e36b37/32.png) [@thomas](https://discuss.vllm.ai/u/thomas)\
**Post date:** [October 4, 2025, 1:21am UTC](https://discuss.vllm.ai/t/which-attention-backend-for-gpt-oss-in-version-0-11-0/1691/1 "2025-10-04T01:21:28Z")

</div>

Which VLLM\_ATTENTION\_BACKEND should I use for gpt-oss in version 0.11.0? I used to have “TRITON\_ATTN\_VLLM\_V1”, but it doesn’t work with 0.11.0 anymore, it shows “Value error, Invalid value ‘TRITON\_ATTN\_VLLM\_V1’“. Please help!

---

<div class="post-metadata">

**Author:** ![RunLLM](https://yyz2.discourse-cdn.com/flex036/user_avatar/discuss.vllm.ai/runllm/32/233_2.png) [@RunLLM](https://discuss.vllm.ai/u/RunLLM)\
**Post date:** [October 4, 2025, 1:22am UTC](https://discuss.vllm.ai/t/which-attention-backend-for-gpt-oss-in-version-0-11-0/1691/2 "2025-10-04T01:22:10Z")

</div>

The value “TRITON\_ATTN\_VLLM\_V1” is not a valid option for VLLM\_ATTENTION\_BACKEND in vLLM 0.11.0. The accepted values are now “FLASH\_ATTN”, “FLASHINFER”, or “XFORMERS”. You should set VLLM\_ATTENTION\_BACKEND to one of these, depending on your hardware and installed libraries. For example:

```bash
export VLLM_ATTENTION_BACKEND=FLASH_ATTN

```

According to the [official documentation](https://docs.vllm.ai/en/latest/getting_started/quickstart.html#on-attention-backends), vLLM will automatically select the best backend if you do not set this variable, but you can override it with one of the supported values.

Would you like more detail on which backend is best for your hardware?

Sources:

- [On Attention Backends (latest)](https://docs.vllm.ai/en/latest/getting_started/quickstart.html#on-attention-backends)
- [On Attention Backends (stable)](https://docs.vllm.ai/en/stable/getting_started/quickstart.html#on-attention-backends)

Hint: Mention @RunLLM in the post for followups.