strands-decider:2b
Strands Decider · Strands Agents · 2B · ONNX Runtime · Apache-2.0
- Score
- 61.3
- 95 % interval 59.3–63.1
- Rank
- 5
- of 41 scored models
- Public benchmark
- 74.5
- macro accuracy, %
- typed-decisions
- 61.6
- accuracy, %
By task
Chance-corrected accuracy on each of the 14 tasks: 0 is guessing, 100 is every answer right.
Accuracy and calibration
| Benchmark | Accuracy | Choice | Noul | Score | ECE | Brier | Rejected |
|---|---|---|---|---|---|---|---|
| Public benchmark | 74.5(73.1–75.7) | 79.6 | 83.4 | 51.2 | 0.039 | 0.347 | 0 / 3,880 |
| typed-decisions | 61.6(59.3–63.9) | 61.0 | 75.0 | 51.9 | 0.090 | 0.521 | 0 / 400 requests |
Accuracy in %, with its 95 % bootstrap interval. A rejected request (the model can’t take the question, or the input is too long for it) counts as wrong.
Speed and memory
| Device | p50 | p90 | Load | VRAM | RAM peak | Precision |
|---|---|---|---|---|---|---|
| RTX 5090 CUDA | 97.6 ms | 100 ms | 31.2 s | 8.5 GB | 11.4 GB | F32 |
| Threadripper 3960X CPU | 4.46 s | 4.51 s | 31.2 s | – | 11.1 GB | F32 |
strands-decider:2b-fp16 RTX 5090 | 71.6 ms | 75.6 ms | 19.1 s | 5.8 GB | 9.9 GB | F16 |
strands-decider:2b-fp16 is the same model computing in fp16 on the GPU, not ranked separately. Its decisions agree with strands-decider:2b's on at least 99.5 % of the parity questions; typed-decisions 61.5 against 61.6, measured with Ollaya 0.14.0.
Five-question requests over HTTP, one at a time, each a different short message. VRAM: the rise of the GPU’s used memory while the model is loaded. RAM peak: the runner process’s peak resident memory.
Parity with the authors’ code
| Device | Same decision | Max logit diff. | Result | Record |
|---|---|---|---|---|
| RTX 4090 CUDA | 650/650 | 1.1e-5 | pass | results/runs/2026-10-19-parity-rtx4090-cuda.json |
Reproduce
The tools are in convert/ollaya_convert and need only Python’s standard library. Run them against a local Ollaya 0.13.0 with the same model digest.
ollaya pull strands-decider:2b
# typed-decisions; export the test split once with
# uv run python -m ollaya_convert.bench_typed export --out typed.jsonl
python -m ollaya_convert.bench_typed run --model strands-decider:2b \
--url http://127.0.0.1:11435 --data typed.jsonl --out typed/strands-decider-2b
# Bespoke Labs' public benchmark; build its 13 subsets first
NIMBLE_SRC=nimble python -m ollaya_convert.bench_public run --model strands-decider:2b \
--url http://127.0.0.1:11435 --data public --out public/strands-decider-2b
# speed and memory
python -m ollaya_convert.bench_latency --models strands-decider:2b --memory \
--machine my-machine --device cuda --out latency.json