CVE-2026-71486

vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the /v1/completions/derender and /v1/chat/completions/derender endpoints accept caller-supplied GenerateResponse objects whose generate_responses, choices, token_ids, prompt_logprobs, logprobs.content, top_logprobs, and routed_experts structures are processed by OnlineDerenderer and tokenizer.decode before max_model_len, max_tokens, max_num_seqs, or response-size limits are enforced, allowing an authenticated API client to consume excessive CPU and memory and produce oversized responses. This issue is fixed in version 0.26.0.

CVSS 3.1

4.3CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:L

EPSS

0.4% chance of exploitation in the next 30 days; higher than 31% of CVEs (scores of 2026-10-25)

CWEs

CWE-400, CWE-770

Published

2026-08-17

Modified

2026-10-23

Record from

the NVD, synced 2026-10-23

Affected

• vllm:vllm

Packages (OSV)

References

• https://github.com/vllm-project/vllm/commit/8e61b646e2d157f9b93451fa048f9c8530c8a67b
• https://github.com/vllm-project/vllm/pull/47260
• https://github.com/vllm-project/vllm/releases/tag/v0.26.0
• https://github.com/vllm-project/vllm/security/advisories/GHSA-8737-qx52-hjff

Check your stackJSON