BBuf/vllm-kimi-optimization
skills/model-optimization/vllm/vllm-kimi-optimization/SKILL.md
PR-backed optimization manual for Kimi K2 / K2.5 / Linear / Audio / VL in vLLM. Use when an engineer needs to audit, debug, extend, or document Kimi-VL, Kimi-Linear, Kimi-K2.5, Kimi-Audio, parser aliases, and quantized MLA behavior in vLLM.
$ install --global
skillsauthnpx skillsauth add BBuf/AI-Infra-Auto-Driven-SKILLS vllm-kimi-optimizationInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: May 2, 2026, 2:40 AM148.1s2 files scanned
SKILL.md
- name:
- vllm-kimi-optimization
- description:
- PR-backed optimization manual for Kimi K2 / K2.5 / Linear / Audio / VL in vLLM. Use when an engineer needs to audit, debug, extend, or document Kimi-VL, Kimi-Linear, Kimi-K2.5, Kimi-Audio, parser aliases, and quantized MLA behavior in vLLM.
vLLM Kimi K2 / K2.5 / Linear / Audio / VL Optimization
Overview
This skill covers Kimi-VL, Kimi-Linear, Kimi-K2.5, Kimi-Audio, parser aliases, and quantized MLA behavior in vLLM.
Evidence snapshot:
- vLLM mainline checked around
0f7be0f2f76814f80f9091220a5fbbb53912ad00 - Support status: supported on current mainline
- Canonical PR notes:
references/pr-history.md - History mirrors:
model-pr-optimization-history/vllm/kimi/README.zh.mdandREADME.en.md
Non-Negotiable Evidence Rule
Use skills/model-optimization/model-pr-diff-dossier/SKILL.md as the production bar.
Every PR cited for this family must be based on diff reading, not only PR titles.
Runtime Surfaces
vllm/vllm/model_executor/models/kimi_vl.pyvllm/vllm/model_executor/models/kimi_linear.pyvllm/vllm/model_executor/models/kimi_k25.pyvllm/vllm/model_executor/models/kimi_audio.py
Current Main Summary
- The Kimi family in vLLM spans vision, linear-attention, K2.5, and audio checkpoints.
- The most fragile areas are MLA plus FP8/NVFP4 loading, processor evolution, and parser alias compatibility between K2 and K2.5.
Key Landed PRs
- #16387
Add Kimi-VL model support: Landed the original Kimi-VL multimodal runtime. - #27809
Introduce Kimi Linear to vLLM: Added the linear-attention Kimi family instead of only the VL path. - #33131
Kimi-K2.5: Brought the K2.5 generation into mainline. - #33876
Fix Kimi-K2.5 NVFP4 checkpoints weight loading: Closed a concrete launch blocker for quantized K2.5 checkpoints. - #36127
Add support for moonshotai/Kimi-Audio-7B-Instruct: Extended the family to audio-conditioned serving. - #37438
Add Kimi-K2.5 reasoning/tool parser aliases: Aligned parser aliases and tool-call IDs with the newer model outputs.
Open Optimization Items
- Re-run PR search before claiming new support beyond the checked mainline commit.
Validation Lanes
- K2.5 text and VL launch.
- Kimi-Linear long-context inference.
- Kimi-Audio processor and transcription path.
- NVFP4 or FP8 MLA startup on K2.5.
Related Skills
BBuf/vllm-sota-humanize-loop
development
VerifiedTrustedCommunity
Run an autonomous Humanize-governed vLLM SOTA performance loop for one LLM model: first perform the fixed fair vLLM/SGLang/TensorRT-LLM deployment search and benchmark, then start one RLCR loop that repeatedly decides the gap, profiles the current bottleneck, runs layer/kernel pipeline analysis, patches vLLM code, optionally uses ncu-report-skill for kernel evidence, and revalidates until vLLM matches or beats the best observed framework under the same workload and SLA.
423SKILL.mdUpdated May 27, 2026
BBuf/llm-pipeline-analysis
devops
VerifiedTrustedCommunity
Inspect LLM torch profiler traces at forward-pass, layer, and kernel level. Use when you need layer timings, anchor-kernel boundaries, representative kernel flows, or Perfetto time ranges.
423SKILL.mdUpdated May 21, 2026
BBuf/sglang-sota-humanize-loop
development
VerifiedTrustedCommunity
Run an autonomous Humanize-governed SGLang SOTA performance loop for one LLM model: first perform the fixed fair SGLang/vLLM/TensorRT-LLM deployment search and benchmark, then start one RLCR loop that repeatedly decides the gap, profiles the current bottleneck, runs layer/kernel pipeline analysis, patches SGLang code, optionally uses ncu-report-skill for kernel evidence, and revalidates until SGLang matches or beats the best observed framework under the same workload and SLA.
423SKILL.mdUpdated May 17, 2026
BBuf/model-pr-history-knowledge
documentation
VerifiedTrustedCommunity
Use when an SGLang, vLLM, or TensorRT-LLM serving/model optimization task needs prior model-family PR evidence. Query and read the PR-driven history docs under model-pr-optimization-history before choosing source paths, fast paths, kernel/fusion ideas, regression risks, or validation lanes.
423SKILL.mdUpdated May 17, 2026