TernaryQuench: open-source ternary quantization trainer for Qwen3 with MLX export
casper_hansen_ · x · 2026-10-05
Developer penk released TernaryQuench, an open-source project giving the complete recipe for building ternary language models: start from an upstream Qwen3 checkpoint, generate agentic calibration traces, train with CAT-Q-style reconstruction, and export for local inference.
- Supports Qwen3 and the text decoders of Qwen3.5/Qwen3.8; weights learn three values per group (−scale, 0, +scale)
- Exporters let you mix trained ternary layers with higher-precision layers in one model
- Full pipeline open-sourced: calibration, training, checkpoint recovery, evaluation
- Ships a ready TernaryQuench Qwen3.8-27B GGUF runnable via ollama run, with MLX export for Apple Silicon
More from Infra
- OpenAI reportedly selling Cerebras-powered Ultrafast inference at ~$200M per megawatt — downingARK · 2026-10-05
- Cross-cloud app-db setup sees 13x latency hit, new networking benchmarks show — DanielLockyer · 2026-10-05
- "Free-range tokens": the viral riff on cage-free compute and batch-size-1 dreams — jwt0625 · 2026-10-05
- dhh: a cheap Beelink mini PC with 8745HS runs local AI setups just fine — viksit · 2026-10-05
- Ex-Google engineer who trained first Gemini Nano sees on-device ML inflection in 2027/28 chips — _arohan_ · 2026-10-05
- Vercel Engineer Calls for Shared Fund to Fix KVM Bugs Affecting All Hyperscalers — cramforce · 2026-10-05