{"id":51027161,"url":"https://github.com/sohanemon/nvidia-nim-benchmark","last_synced_at":"2026-06-21T20:30:42.598Z","repository":{"id":355741526,"uuid":"1229402753","full_name":"sohanemon/nvidia-nim-benchmark","owner":"sohanemon","description":"Benchmark latency of NVIDIA NIM models (LLMs, inference speed test tool)","archived":false,"fork":false,"pushed_at":"2026-05-05T02:27:35.000Z","size":3,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-05-05T04:30:23.727Z","etag":null,"topics":["benchmark","llm","nim","nvidia","python"],"latest_commit_sha":null,"homepage":"https://sohanjs.web.app/readme/nvidia-nim-benchmark","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/sohanemon.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-05-05T02:22:00.000Z","updated_at":"2026-05-05T02:29:23.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/sohanemon/nvidia-nim-benchmark","commit_stats":null,"previous_names":["sohanemon/nvidia-nim-benchmark"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/sohanemon/nvidia-nim-benchmark","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sohanemon%2Fnvidia-nim-benchmark","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sohanemon%2Fnvidia-nim-benchmark/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sohanemon%2Fnvidia-nim-benchmark/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sohanemon%2Fnvidia-nim-benchmark/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/sohanemon","download_url":"https://codeload.github.com/sohanemon/nvidia-nim-benchmark/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/sohanemon%2Fnvidia-nim-benchmark/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34625624,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-21T02:00:05.568Z","response_time":54,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["benchmark","llm","nim","nvidia","python"],"created_at":"2026-06-21T20:30:41.508Z","updated_at":"2026-06-21T20:30:42.580Z","avatar_url":"https://github.com/sohanemon.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# NVIDIA NIM Speed Benchmark\n\nAuto-discovers available chat models from your NVIDIA API key and benchmarks them for latency and throughput.\n\n## Requirements\n\n- Python 3.8+\n- NVIDIA API key ([get one here](https://org.ngc.nvidia.com/))\n\n## Setup\n\n```bash\n# Set your API key\nexport NVIDIA_API_KEY=your_key_here\n\n# Run the benchmark\npython3 main.py\n```\n\n## Usage\n\n```bash\n# Basic usage\npython3 main.py\n\n# Test only top 5 fastest models\npython3 main.py --top 5\n\n# Run multiple iterations per model for averaging\npython3 main.py --runs 3\n\n# Filter to specific model families (e.g., llama, mistral)\npython3 main.py --filter llama\n```\n\n### Options\n\n| Flag | Description | Default |\n|------|-------------|---------|\n| `--top N` | Only test first N models | All |\n| `--runs N` | Runs per model for averaging | 1 |\n| `--filter STR` | Only test models containing STR | None |\n\n## Output\n\nThe benchmark outputs:\n\n- **Latency** — Time to first token (seconds)\n- **Tokens** — Completion tokens generated\n- **tok/s** — Throughput (tokens/second)\n- **Leaderboard** — Ranked fastest to slowest\n\n### Example Output\n\n```\nFetching model list from API...\nFound 12 model(s) to test\n\n======================================================================\n  NVIDIA NIM Speed Benchmark   (runs/model: 1)\n======================================================================\n\n  \u003e\u003e llama-3.1-405b-instruct                     2.31s | 58 tok | 25.1 tok/s\n  \u003e\u003e llama-3.1-70b-instruct                      1.89s | 57 tok | 30.2 tok/s\n  \u003e\u003e mixtral-8x7b-instruct                       1.12s | 56 tok | 50.0 tok/s\n\n======================================================================\n  LEADERBOARD  (fastest to slowest)\n======================================================================\n  #    Model                                             Latency    tok/s\n  --------------------------------------------------------------------\n  1st  mixtral-8x7b-instruct                            1.12s      50\n  2nd  llama-3.1-70b-instruct                           1.89s      30\n  3rd  llama-3.1-405b-instruct                          2.31s      25\n\n  WINNER  : mixtral-8x7b-instruct\n  Latency : 1.12s\n  tok/s   : 50\n  Reply   : \"Space exploration has revolutionized our understanding of the universe.\"\n```\n\n## How It Works\n\n1. Fetches all available chat models from `https://integrate.api.nvidia.com/v1/models`\n2. Sends a simple prompt to each model (\"Reply with exactly one short sentence about space.\")\n3. Measures latency from request to first response\n4. Calculates tokens/second throughput\n5. Ranks models by latency\n\n## Notes\n\n- Models that aren't chat-compatible or lack API access are automatically skipped\n- The `--runs` flag helps smooth out variance from network latency\n- Use `--filter` to focus on specific model families you have access to","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsohanemon%2Fnvidia-nim-benchmark","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsohanemon%2Fnvidia-nim-benchmark","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsohanemon%2Fnvidia-nim-benchmark/lists"}