Strands Decider — local AI model by AWS Strands Labs
Written by Jakub Rusinowski · Last updated
AWS's open decision model: a Qwen3.5-2B-Base backbone with its language-model head removed, a LoRA and a small readout head that scores the options of a typed question. Apache-2.0, installed with pip and served by strands-decider serve at /v1/systemone.
Variants
The smallest Strands Decider variant needs about 2 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache.
| Model | VRAM at Q4 | VRAM | Context | Run it |
|---|---|---|---|---|
| Strands Decider 2B (hobson-v21) → 2B | ~1.9 GB | Not published | strands-decider serve :8000 |
Memory is quantized weights plus overhead at Q4_K_M, from the same engine as the GPU & VRAM checker.
How to run Strands Decider locally
Install Ollama, then pull the tag.
Served by strands-decider serve (default port 8000). This model does not run in Ollama.
Pick a size above for its own VRAM figure, speed estimate and install command.
Licence
Commercial use permitted. No usage restrictions beyond attribution.
Applies to: Strands Decider 2B (hobson-v21)Recommended GPU
The cheapest catalogued GPU that runs Strands Decider locally (min 2 GB VRAM) is the Intel Arc B570 (10 GB).
Strands Decider — frequently asked questions
How much VRAM does Strands Decider need?
Strands Decider needs about 2 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Strands Decider 2B (hobson-v21) (2 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Strands Decider on an RTX 4090 (24 GB)?
Yes — Strands Decider runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for Strands Decider?
Q4_K_M is the best balance of quality and VRAM for Strands Decider in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run Strands Decider locally?
Strands Decider is a decision model: it is called over an HTTP endpoint (/v1/systemone), not chatted with. Where a variant is available in Ollama 0.35 or newer, pull it with `ollama pull` and send requests to the local server; the others ship their own server. Each variant page shows the exact commands for that model.