PurpleMIST-Flash-1.0 (Olaverse)
Data: JevBench v1.6.39, published 2026-10-22
Capability 60.1 · Not ranked in the current release
unpriced: the frozen pricing path has no market-listing decision for the declared base Qwen/Qwen3.5-9B-Base and raises rather than estimating; one listing decision would make the row rankable.
Last measured: 2026-10-17 · measurement release: v1.6.28
Apache-2.0 (model card and HF metadata), public, not gated
intelligence
36.6
calibration
83.6
speed
88.3
cost
—
Composite (secondary)
— · Not ranked on this board
Cost and measurement conditions
not published (not published)
UNPRICED. The frozen pricing_v15 module (sha256 6202ef78..., the module in force at scoring time; the later re-freezes only add or re-price hosted-API launch-list keys) has no market-listing decision for the declared base Qwen/Qwen3.5-9B-Base (the module RAISES 'base model lacks an explicit market-listing decision', for every spelling tried); no price is invented here. Axes only: Intelligence, Calibration and Speed are computed; Cost, composite and rank are void - the same treatment as the live jevora-9b-v1 and kev-9b-v2 rows on that same declared base. The tokens needed for a later reprice are measured and on file: the author's server reports NO usage block at all, so the input tokens were counted OFFLINE with the author's own pinned tokenizer and their own unchanged encode() on the exact prompts their server builds from the runner's request bodies (CoS deputy decision #18974, as accepted for the sibling row): 700,492 input tokens over the 1,485 answered items, 0 output (nothing is generated). That count is byte-for-byte identical, item by item, to the count accepted for purplemist-mini-1.0 - same rendered-prompt-ids sha256 ad0810696b7e80eba87192caa70f57c19cdbee103e37f3d9cd163ef5df434c86 - which it must be, because the rendered prompt depends only on the tokenizer, encode(), the two flags and the request body, all byte-identical between the two rows, and not on the weights.
Measured p50 latency: 0.084 s (on-GPU, no network, loopback; not directly comparable with API latency, which we time end to end from our measuring host, network included)
Ranking and cap use: 0.32 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.084 s on-GPU, no network (timed on our GPU over loopback). Not directly comparable with API latency, which we time end to end from our measuring host, network included. The ranking and latency cap use the same estimate for every self-hosted row: 2 × measured + 150 ms = 0.32 s. Model authors can ask us to time a hosted endpoint instead (on request).
x2 + 0.15 s (assumption, not measured)
Evaluator-owned rented Lium GPU pod, 1x NVIDIA RTX 6000 Ada Generation 48 GB (sm_89, driver 595.84), Athens, Greece (EU), secure tier (not preemptible), USD 0.80/h, 05:26-05:53Z on 17 Oct 2026 = USD 0.35; removed by exact id and verified gone four ways. Speed is as measured on that card and is NOT strictly comparable to the H100/RTX 5090 rows. Isolation: the author's server ran inside a network namespace with loopback only, with the cut PROVEN before any sealed item was dispatched (no interfaces but lo, no routes, outbound by name and by IP both failing) and proven again immediately before the first sealed request; credential-free environment (env -i, checked), weights from a locally verified offline snapshot of the pinned revision (11 files, 15,927,059,974 bytes, model.safetensors sha256 2dba83efd90dbe2fcf767eaa004f48789029cff3282f7c6ae8a66eca7dc26202), sealed input deleted on the pod at 05:51:55Z and the server's process group killed by exact pid. One request at a time. Transfer deviation: the Hugging Face Xet path stalled on this host (its own log reported 'connection struggling' at a success ratio of 1.000 while the link was taking 25 MB/s), so the snapshot was fetched over plain HTTPS with HF_HUB_DISABLE_XET=1; what is verified did not change.