{"id":"CVE-2026-100653","aliases":["EUVD-2026-87747","GHSA-hhv2-872h-628q","GHSA-r3cf-2wgr-7xh9","PYSEC-2026-4324"],"status":"published","title":"vLLM 0.22.1 before 0.28.0 Incomplete Artifact Pin Propagation","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","severity":{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-10","score":0.0034,"percentile":0.27663,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-348"],"published":"2026-09-26T13:23:22Z","modified":"2026-10-28T02:30:27Z","fixes":[{"ecosystem":"PyPI","package":"vllm","fixed_in":["0.28.0"],"sources":["osv"]}],"affected":[{"source":"osv","ecosystem":"PyPI","package":"vllm","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0.22.1"},{"fixed":"0.28.0"}]}],"versions":["0.22.1","0.23.0","0.24.0","0.25.0","0.25.1","0.26.0","0.27.0","0.27.1"]}],"products":[{"vendor":"vllm","product":"vllm"},{"vendor":"vllm-project","product":"vllm"}],"references":[{"url":"https://github.com/vllm-project/vllm/security/advisories/GHSA-hhv2-872h-628q","tags":["vendor-advisory","exploit","Exploit","Mitigation","Vendor Advisory","advisory","evidence"]},{"url":"https://github.com/vllm-project/vllm/commit/d26a28ab033697f55a1414b5b0435de7cd6045b6","tags":["patch","Patch","fix"]},{"url":"https://www.vulncheck.com/advisories/vllm-0.22.1-before-0.28.0-incomplete-artifact-pin-propagation","tags":["third-party-advisory","Third Party Advisory","advisory"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-100653","tags":["advisory"]},{"url":"https://github.com/advisories/GHSA-r3cf-2wgr-7xh9"},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/100xxx/CVE-2026-100653.json","tags":["advisory"]}],"sources":{"cve":[{"id":"CVE-2026-100653","title":"vLLM 0.22.1 before 0.28.0 Incomplete Artifact Pin Propagation","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","status":"published","published":"2026-09-26T13:23:22.605Z","modified":"2026-10-01T11:06:45.139Z","cvss":[{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high","assessor":"cna"},{"version":"3.1","score":6.5,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:H/A:N","level":"medium","assessor":"cna"}],"cwes":["CWE-348"],"updated_at":"2026-11-02T14:20:06.925367Z"}],"euvd":[{"id":"EUVD-2026-87747","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","status":"published","published":"2026-09-26T13:23:22Z","modified":"2026-10-01T11:06:45Z","cvss":[{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high"}],"updated_at":"2026-11-02T17:27:46.486382Z"}],"ghsa":[{"id":"GHSA-r3cf-2wgr-7xh9","title":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1...","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","status":"published","published":"2026-09-26T15:31:18Z","modified":"2026-09-26T15:31:23Z","cvss":[{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high"},{"version":"3.1","score":6.5,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:H/A:N","level":"medium"}],"cwes":["CWE-348"],"updated_at":"2026-11-02T15:12:30.047516Z"}],"nvd":[{"id":"CVE-2026-100653","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","status":"published","published":"2026-09-26T14:16:47.953Z","modified":"2026-10-20T18:35:50.567Z","cvss":[{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high","assessor":"secondary"},{"version":"3.1","score":6.5,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:H/A:N","level":"medium","assessor":"secondary"}],"cwes":["CWE-348"],"updated_at":"2026-11-02T14:12:52.081323Z"}],"osv":[{"id":"CVE-2026-100653","title":"vLLM 0.22.1 before 0.28.0 Incomplete Artifact Pin Propagation","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","status":"published","published":"2026-09-26T13:23:22.605Z","modified":"2026-10-28T02:30:27.428912653Z","cvss":[{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N","level":"high"}],"cwes":["CWE-348"],"updated_at":"2026-11-02T15:44:53.490346Z"},{"id":"PYSEC-2026-4324","description":"vLLM is an inference and serving engine for large language models. In versions from 0.22.1 through 0.28.0, the operator-supplied model revision pin (--revision / --code-revision) is not propagated to several Hugging Face artifact loads for the FunAudioChat and Tarsier2 architectures: the WhisperFeatureExtractor and speech_tokenizer PreTrainedTokenizerFast loads in vllm/model_executor/models/funaudiochat.py and the Qwen2VLConfig.from_pretrained call used by Tarsier2ProcessingInfo in vllm/model_executor/models/qwen2_vl.py. As a result, deployments pinned to a reviewed revision still resolve these behavior-affecting processor, tokenizer, and config artifacts from the repository's default revision, so a later change to the upstream default branch can alter audio preprocessing, speech tokenizer behavior, or Tarsier2 configuration without any change to the operator's configured pin. This is a supply-chain integrity and reproducibility failure for pinned deployments; it is residual to the earlier fix tracked as GHSA-3ww4-5jv9-j5gm / CVE-2026-47155 and does not constitute remote code execution or a trust_remote_code=False bypass. The issue is fixed in version 0.28.0.","status":"published","published":"2026-09-26T14:16:47.953Z","modified":"2026-10-23T19:00:39.873437407Z","cvss":[{"version":"4.0","score":8.3,"vector":"CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:L/VI:H/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X","level":"high"}],"updated_at":"2026-11-02T15:52:57.671918Z"}]}}