Show HN: Timber – Ollama for classical ML models, 336x faster than Python

Hacker News · Mar 2, 2026 · Collected from RSS

Full Article

Timber Ollama for classical ML models. Timber compiles trained tree-based models (XGBoost, LightGBM, scikit-learn, CatBoost, ONNX) into optimized native C and serves them over a local HTTP API. No Python runtime in the inference hot path Native latency (microseconds) One command to load, one command to serve 📚 Docs: https://kossisoroyce.github.io/timber/ Who is this for? Timber is built for teams that need fast, predictable, portable inference: Fraud/risk teams running classical models in low-latency transaction paths Edge/IoT teams deploying models to gateways and embedded devices Regulated industries (finance, healthcare, automotive) needing deterministic artifacts and audit trails Platform/infra teams replacing Python model-serving overhead with native binaries Quick Start pip install timber-compiler # Load any supported model (auto-detected) timber load model.json --name fraud-detector # Serve it (Ollama-style workflow) timber serve fraud-detector curl http://localhost:11434/api/predict \ -d '{"model": "fraud-detector", "inputs": [[1.0, 2.0, 3.0, ...]]}' Supported Formats Format Framework File Types XGBoost JSON XGBoost .json LightGBM text LightGBM .txt, .model, .lgb scikit-learn pickle scikit-learn .pkl, .pickle ONNX ML opset (TreeEnsemble) ONNX .onnx CatBoost JSON CatBoost .json Benchmarks (Methodology + Reproducibility) The 336× claim is measured against Python XGBoost single-sample inference. Methodology Hardware: Apple M2 Pro, 16 GB RAM, macOS (recorded by script) Model: XGBoost binary classifier, 50 trees, max depth 4, 30 features Dataset: breast_cancer (sklearn) Warmup: 1,000 iterations Timed: 10,000 single-sample predictions Metric: in-process latency (not HTTP/network round-trip) Baseline: Python XGBoost (booster.predict) Reproducible scripts See benchmarks/ for: run_benchmarks.py (Timber vs Python XGBoost + optional ONNX Runtime/Treelite/lleaves) system_info.py (hardware/software metadata) render_table.py (markdown table output) Run: python benchmarks/run_benchmarks.py --output benchmarks/results.json python benchmarks/render_table.py --input benchmarks/results.json Comparisons Runtime Runtime deps Typical artifact size Latency profile Notes Timber None (generated C99) ~48 KB (example model) ~2 µs native call Strong fit for edge/embedded and deterministic deployments Python (xgboost/sklearn serving) Python + framework stack 50–200+ MB process footprint 100s of µs to ms Easy dev loop, high runtime overhead ONNX Runtime ONNX Runtime libs MBs to 10s of MBs usually low 100s of µs Broad model ecosystem, larger runtime Treelite Runtime Treelite runtime + compiled artifact MB-scale runtime + model lib low-latency when compiled Great for GBDTs; separate compile/runtime flow lleaves Python package + LightGBM text model Python runtime + compiled code lower than pure Python LightGBM-focused Limitations / Known Issues ONNX support is currently focused on TreeEnsembleClassifier/Regressor operators. CatBoost support expects JSON exports (not native binary formats). scikit-learn parser supports major tree estimators and pipelines; uncommon/custom estimator wrappers may fail. Pickle parsing follows Python pickle semantics — only load trusted artifacts. XGBoost support is JSON-model based. Binary booster formats are not the primary input path. Optional benchmark backends (ONNX Runtime, Treelite, lleaves) are skipped unless installed/configured. API Endpoints (serve mode) Endpoint Method Description /api/predict POST Run inference /api/generate POST Alias for /api/predict (Ollama compat) /api/models GET List loaded models /api/model/:name GET Get model metadata /api/health GET Health check Roadmap Improve framework/version compatibility coverage (including more edge-case model exports) Broaden ONNX operator support beyond tree ensembles Strengthen embedded deployment profiles (ARM Cortex-M / RISC-V presets) Add richer benchmark matrices and public reproducibility reports Expand safety/regulatory tooling around audit + MISRA-C workflows Examples End-to-end runnable examples live in examples/: quickstart_xgboost.py quickstart_lightgbm.py quickstart_sklearn.py They generate model files you can load immediately with timber load. Paper Timber includes a full technical paper: paper/timber_paper.pdf Citation (BibTeX) @misc{royce2026timber, title = {Timber: Compiling Classical Machine Learning Models to Native Inference Binaries}, author = {Kossiso Royce}, year = {2026}, howpublished = {GitHub repository and technical paper}, institution = {Electricsheep Africa}, url = {https://github.com/kossisoroyce/timber} } Community & Governance Contributing guide: CONTRIBUTING.md Code of conduct: CODE_OF_CONDUCT.md Security policy: SECURITY.md Development pip install -e ".[dev]" pytest tests/ -v License Apache-2.0