StartLux

We’ve been shipping. Here’s what’s out. 🛠️ StartLux-27B A 27B model for tool use and multi-step task execution. In CAICT’s August 2026 MCP evaluation, StartLux-V1.0-27B-Preview scored 39.25% Overall Micro Pass@1, ranking #2 among six tested models. ⚡ StartLux-Decision Fast decisions when confident. Deeper reasoning when needed. Model weights and inference code for the StartLux-Decision family are available. On October 16, StartLux-Decision-v1.1 27B reached #1 on the Hugging Face Jev Decision Index 0.3.2 text leaderboard. 🔧 StartLux-Prism To make decision models easier to use and deploy, we released and open-sourced an inference engine built specifically for StartLux-Decision, delivering up to 5.3× speedup on a single GPU. Faster decision responses. More efficient computation. 🧭 DecisionPivot A benchmark for evaluating decision-making at key points in LLM systems: whether decisions are correct, whether models still make the right decision when a decisive detail changes, and how much work they can automate at a fixed error rate.

⚡️ Same model. Same GPU. Up to 5.3× the speed. Today we’re open-sourcing StartLux-Prism, our dedicated inference engine for decision models. On a single H200, running StartLux-Decision v1.1 27B AutoThink, Prism INT8 delivers 1.2–5.3× the speed of our vLLM 0.31 FP8 setup across four tested workloads. • Short question (~256 tokens): 53.1 → 19.0 ms — 2.8× speedup. • Three questions sharing one context: 92.9 → 17.4 ms — 5.3×. • The same three questions with reasoning: 1,296 → 785 ms — 1.7×. • Long question (~8,192 tokens): 531.2 → 439.8 ms — 1.2×. Accuracy with reasoning stays nearly unchanged: 92.94% vs. 92.97% with BF16 across seven benchmarks covering 12,351 decisions. Results stay reproducible, too. With fixed hardware and weights, output probabilities remain bit-identical across tested batching and caching configurations. ✨ From models to inference, we’re expanding the local AI stack. Code: github.com/StartLuxLabs/Star… Models: Hugging Face: huggingface.co/startlux-mode… ModelScope:modelscope.cn/organization/S… API: bamboonode.dev

Most models generate. Decision is harder. Today we’re releasing the latest StartLux-Decision models: a family of decision models built for fast choices and deeper reasoning. • Instant: single-pass decisions, from 0.8B to 27B • AutoThink: adaptive reasoning, from 4B to 27B #1 among evaluated models on the public subset of Decision Index 0.3. Our AutoThink 27B scores 73.37, outperforming Jev (hosted) by 15.41 points. Model weights and inference code are now available: github.com/StartLuxLabs/Star… Free API testing of AutoThink 27B is also available through a third-party platform: bamboonode.dev/

Introducing StartLux-Decision — a family of open-weight models for fast, local decision-making. · Decision performance: 27B scores 63.88 on Decision Index 0.2.1, ahead of Jev 1.13 on 31 out of 38 benchmarks. · Accuracy: 27B averages 91.82% across the seven test suites used by Intern-Decision, compared with Jev’s 88.74% and Intern-Decision-4B’s 90.02%. · Speed: In our short-request test, 4B handled three decisions in a single request, averaging 26 ms.* · Local deployment: Five dense sizes (0.8B–27B), with original and GGUF weights. On Decision Index 0.2.1, our 0.8B, 2B, 4B, and 9B models exceed the highest public leaderboard score at each respective size; 27B exceeds the overall leader, Jev 1.13. Models: huggingface.co/collections/s… Code & results: github.com/StartLuxLabs/Star… Self-evaluated results; Jev comparison uses the September 28, 2026 public leaderboard. *Short-request test on one H200, BF16, with fast kernels and CUDA graphs enabled.