- Services
StartLux-Decision v1.1 27B AutoThink is an open-weights decision model from StartLux. It takes a state as text, JSON or up to 8 images and answers typed questions: pick one of up to 26 options, yes or no, or a level on a scale of 2 to 10. It reads a probability for every option in one pass and writes a short thought only when the top probability falls below the gate, 0.6 by default. Weights are released under CC BY-NC 4.0 in BF16, GGUF and MLX formats with a 65,536 token context.
A workable 27.36B-parameter dense decision model from startlux-models. Treat the modality benchmarks above as the leading indicator of fit — composite scoring across modalities is still maturing.
Generated from this model’s benchmarks and ranking signals. Editor reviews refine it over time.
Access model weights, configuration files, and documentation.
The top devices for this model at 4-bit, ranked by fit and speed.
| Device | Grade | VRAM |
|---|---|---|
| Acer Veriton GN100 AI MiniAcer | SS | 30.7 GB |
| AI TOP ATOM ATAGB10-9003Gigabyte | SS | 30.7 GB |
| AMD Instinct MI300XAMD | SS | 30.7 GB |
| AMD Instinct MI325XAMD | SS | 30.7 GB |
| AMD Instinct MI355XAMD | SS | 30.7 GB |
Cheapest current cloud rentals with at least 31 GB VRAM, refreshed hourly.
Advertising disclosure: we earn commissions when you shop through the links below.
| Option | Cost / GPU-hour |
|---|---|
NVIDIA RTX A6000Vast.ai · Spot · 48 GB VRAM | $0.31 |
NVIDIA GeForce RTX 5090Vast.ai · Spot · 32 GB VRAM | $0.33 |
NVIDIA RTX A6000RunPod · Community · 48 GB VRAM | $0.33 |
NVIDIA L40Vast.ai · Spot · 48 GB VRAM | $0.37 |
NVIDIA RTX PRO 5000 BlackwellVast.ai · Spot · 48 GB VRAM | $0.40 |
Per-GPU rate across RunPod, the Vast.ai marketplace, DigitalOcean, and Vultr.
Spot tier is interruptible. Plan for restarts when comparing against on-demand prices.
StartLux-Decision v1.1 27B AutoThink is part of the StartLux-Decision family of open decision models, which spans 0.8B to 27B in Instant and AutoThink variants. Each typed question is read as its own prompt holding the whole state; the model reads the option letters in one pass and, if the top probability is below the gate, writes a thought of at most 4,096 tokens and mixes the two read-outs at 0.25 and 0.75. Requests and responses follow the System One /v1/systemone format, and the engine supports vLLM, llama.cpp and MLX. On the public subset of Decision Index 0.3 it scores 73.37 at gate 0.9, and it reaches 86.9% pair accuracy on DecisionPivot at gate 0.9.
Explore the Provider
Aggregate stats, leaderboard, release timeline, and benchmark coverage across every startlux-models model we track.
Check which of your devices can run this model and how fast.
Compare hosted per-token prices before you commit to local hardware.
Break-even math for self-hosting, renting a GPU, or paying per token.
Live status for the hosted APIs you might use instead of running this locally.
View on Hugging Face