snap:2b-q4_k_m
snap · logitlab · 2B · llama.cpp · Apache-2.0
- Score
- 54.4
- 95 % interval 52.3–56.4
- Rank
- 16
- of 35 scored models
- Public benchmark
- 69.1
- macro accuracy, %
- typed-decisions
- 65.5
- 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 | 69.1(67.7–70.4) | 69.0 | 80.5 | 50.0 | 0.043 | 0.415 | 0 / 3,880 |
| typed-decisions | 65.5(63.3–68.0) | 64.8 | 76.5 | 57.9 | 0.043 | 0.476 | 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 | 78.2 ms | 78.7 ms | 4.10 s | 2.5 GB | 1.8 GB | Q4_K - MEDIUM |
| Threadripper 3960X CPU | 1.73 s | 1.76 s | 5.51 s | – | 2.7 GB | Q4_K - MEDIUM |
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
Its parity check is in the model’s family documentation in the repository (docs/families/); the compiled parity file does not list it yet.
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 snap:2b-q4_k_m
# 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 snap:2b-q4_k_m \
--url http://127.0.0.1:11435 --data typed.jsonl --out typed/snap-2b-q4_k_m
# Bespoke Labs' public benchmark; build its 13 subsets first
NIMBLE_SRC=nimble python -m ollaya_convert.bench_public run --model snap:2b-q4_k_m \
--url http://127.0.0.1:11435 --data public --out public/snap-2b-q4_k_m
# speed and memory
python -m ollaya_convert.bench_latency --models snap:2b-q4_k_m --memory \
--machine my-machine --device cuda --out latency.json