{"id":"be7c4b90-e9f7-47ca-a312-7856a1acf3b0","slug":"rigbase-data","kind":"mcp","name":"rigbase-data","summary":"Measured local-LLM tok/s on home hardware (CC BY 4.0) - from getrigbase.com","owner":"getrigbase","repoUrl":"https://github.com/getrigbase/rigbase-data","license":"CC-BY-4.0","licenseTier":"permissive","trust":"community","stars":0,"match":null,"fewStars":true,"format":"mcp","createdAt":"2026-10-28T04:30:06.000Z","addedAt":"2026-11-08T04:41:57.913Z","logoUrl":"https://avatars.githubusercontent.com/u/347532506?v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["data-analytics"],"tags":["ai-hardware","dataset","gpu","homelab","llama-cpp","llm-benchmark","local-llm","vllm"],"runtime":"remote","local":false,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-11-08T04:44:34.025Z","sample":false,"url":"https://conduits.dev/tools/rigbase-data","language":"","transports":["http"],"install":{"remoteUrl":"https://getrigbase.com/mcp"},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# RigBase data\n\n**Measured local-LLM speeds on real home hardware.**\n\n[getrigbase.com](https://getrigbase.com) · [Live download](https://getrigbase.com/data/) · [Hugging Face](https://huggingface.co/datasets/rigbase/rigbase-local-llm-benchmarks) · Licence: [CC BY 4.0](LICENSE)\n\nThis repository mirrors the free RigBase dataset:\n\n| | |\n|---|---|\n| Builds | **150** |\n| Measured results | **358** |\n| Models | **51** |\n| Snapshot generated | **2026-10-27** (from getrigbase.com/data/) |\n\nEvery number links to its original public source (llama.cpp discussions, llama-bench databases, blogs, GitHub repos, Mia's AI Lab recipes, and similar). RigBase collects and normalises these numbers; it does **not** re-measure them or invent estimates for the `tps` column.\n\n## Files\n\n- `builds.csv` — one row per measured result (358 rows)\n- `builds.json` — one object per build with parts + nested results\n- `LICENSE` — CC BY 4.0\n- `CITATION.cff` — machine-readable citation\n\n## Quick start\n\n```bash\n# CSV\ncurl -LO https://getrigbase.com/data/builds.csv\n\n# Or clone this repo\ngit clone https://github.com/getrigbase/rigbase-data.git\n```\n\n```python\nimport pandas as pd\ndf = pd.read_csv(\"builds.csv\")\n# Plain decode only (drop speculative / MTP / draft runs):\nplain = df[df[\"spec_decode\"].fillna(\"none\").eq(\"none\")]\nprint(plain.groupby(\"model\")[\"tps\"].max().sort_values(ascending=False).head(10))\n```\n\n## Columns (CSV)\n\n| column | meaning |\n|---|---|\n| `build_slug`, `build_title`, `page_url` | Build id, name, and RigBase page |\n| `gpus`, `vram_gb`, `cpu`, `ram`, `power` | Hardware as stated by the source |\n| `model`, `quant`, `engine` | What was run, as stated by the source |\n| `tps` | Generation (decode) tokens/sec reported by the source |\n| `spec_decode` | `none`, or the speculative-decoding method (MTP, DFlash, …) |\n| `prefill_tps`, `prefill_tokens`, `prefill_metric` | Prefill speed only where the source reports one; never estimated |\n| `est_ttft_8k_s_calculated` | Calculated (not measured): 8192 / prefill_tps when prefill_tokens ≥ 512 |\n| `ttft_s_measured`, `ttft_prompt_tokens` | TTFT as measured by the source |\n| `source` | Original public measurement URL |\n| `source_checked_by_rigbase` | `yes` = RigBase re-read the number at the source |\n\nFull column notes: https://getrigbase.com/data/\n\n## Caveats\n\n- Numbers come from different people, engines, quants and settings. Compare rows from the same source when you compute ratios.\n- Filter on `spec_decode == \"none\"` for plain decode speed.\n- Calculator pages on the site (VRAM, KV cache) are **arithmetic**, not measurements. They are not in this dataset.\n\n## Related\n\n- Browse builds: https://getrigbase.com/\n- Can I run it? (VRAM + KV cache calculator): https://getrigbase.com/can-it-run/\n- DGX Spark recipes: https://getrigbase.com/dgx-spark/\n- Submit a correction or build: https://getrigbase.com/submit/\n\n## Attribution\n\nRequired by CC BY 4.0:\n\n> RigBase (getrigbase.com), CC BY 4.0. Each row keeps its original source URL.\n\nPlease also keep or cite the per-row `source` URL when you reuse an individual number.\n\n## Updates\n\nThe live files at https://getrigbase.com/data/ are the source of truth. This GitHub mirror is refreshed when the site dataset changes. Issues and PRs that fix a sourced number (with a link) are welcome.\n"}