llm-inference-benchmark SKILL.md skillA
llm-inference-benchmark from soulmachine/skills is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add soulmachine/skills/llm-inference-benchmark@git:20260613.d71ed01Pins git:20260613.d71ed01 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- soulmachine/skills
- Path
llm-inference-benchmark/SKILL.md- Versions recorded
- 2
- Latest version
- git:20260613.d71ed01, committed Jun 13, 2026
- First version recorded
- Jun 12, 2026
- Size
- 11,214 bytes
- Repository stars
- 3
- License
- MIT
Benchmark OpenAI-compatible LLM inference servers (vLLM, SGLang, or anything serving /v1/completions; local, cross-host via TARGET_HOST=<peer LAN IP>, or behind a TLS+Bearer proxy via --base-url + OPENAI_API_KEY) with sglang.bench_serving run as a standalone dockerized client — one engine-agnostic script (scripts/bench_sweep.sh), default sweep c1→c128 uncapped, sustained-load grid (uniform prompts-per-concurrency), self-describing PROVENANCE log headers — collecting TTFT, TPOT, ITL, and…
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# Benchmark OpenAI-compatible LLM inference servers One tool, one methodology, engine-agnostic: **`sglang.bench_serving`** as a pure HTTP load generator, always hitting the **OpenAI `/v1/completions`** endpoint (every serious engine serves it), always running as a **standalone dockerized client in its own net namespace** — never `docker exec` into the server, never `--network host`. The server's identity (engine, quant, flags) is *measured into the log*, not assumed: every run opens with a PROVENANCE header. **The point of a sweep is the saturation curve's two knees.** Read them off a wide concurrency sweep on a non-binding server: the **min knee** (highest concurrency where TTFT/TPOT is still flat — the latency-optimal cap) and the **max knee** (where output throughput goes flat — the throughput ceiling). A production concurrency cap belongs between them — see **Finding the saturation knees** below. The reference dataset (Kimi-K2.6 on 8× RTX PRO 6000 Blackwell SE, deployed by `deploy-kimi-k26-on-rtx-pro-6000`) lives in [REFERENCE.md](REFERENCE.md) — use it as the comparison anchor when re-benchmarking that hardware after an image bump, config change, or engine swap. …
Read the whole file at its exact version.
How to install
mdr add soulmachine/skills/llm-inference-benchmark@sha256:eb70efe6e7bf94c3The line at the top of this page pins the label git:20260613.d71ed01, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_tebjvpikqjwoyy6t)
3 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 2 versions of this file, between Jun 12, 2026 and Jun 13, 2026, more than one a day.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20260613.d71ed01 latest | 2026-06-13 | d71ed01 | 11,214 B | A | view · diff |
| git:20260612.204f18b | 2026-06-12 | 204f18b | 5,733 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (11214 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
soulmachine/skills · 3 stars · license MIT · pushed 2026-10-31 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_tebjvpikqjwoyy6t GET https://markdownregistry.com/api/v1/resolve?ref=soulmachine/skills/llm-inference-benchmark GET https://markdownregistry.com/api/v1/blob/eb70efe6e7bf94c31931ab134120208450ac5fce0bbaa6bdca40559ca184d88d
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.