Bespoke open-sources Nimble: contrastive data curation recipe clones TypeSafe's Jev
gowthami_s · x · 2026-09-19
Bespoke Labs released Bespoke Nimble, a fully open (data + model + recipe) replication of TypeSafe's Jev-style fast typed-decision model.
- Data: a new "contrastive data curation" recipe — slightly altering facts to create negative examples pushes the model to discriminate better and calibrate implicitly, eliminating the need for probability labels; 10 fully synthetic categories split into train/eval.
- Training: LoRA finetune of Qwen3.5-9B, distillation-free (Jev used only for evaluation), no RL yet.
- Results: the post-trained model substantially outperforms the base Qwen (66%) on the curated eval.
- Serving: parallel constrained decoding; one-step decisions with no reasoning chain for blazing-fast inference.
Related event: Bespoke open-sources Nimble, an open replica of Jev decision model(2 posts)→
More from coding & agent
- Butterfly traces multi-agent failures with graphs, 62.5% accuracy vs 53.8% for log prompting — JeremyCMorgan · 2026-11-04
- Free course covers the 4 hard problems of production generative media agents — Al_Grigor · 2026-11-04
- Quota reset night: Codex back to 100% while Claude shows just 10% left — rudrank · 2026-11-04
- Self-improving software is inevitable: closing the agent-observability loop — bibryam · 2026-11-04
- Building a 7-GPU 132GB VRAM local inference rig: consolidate on EPYC or split across two systems? — infectiousstupidity · 2026-11-04
- Founder of 50-person, 12-country team says he talks to AI agents more than humans — Scobleizer · 2026-11-04