Late entry, not in standings · League 1 field

REAX 1.3

Published

Results f24d1f41, updated 16 November 2026, 21:24 UTC; scored by GNU Backgammon 1.08, 2-ply.

Late entry, not in standings

Late entry
Late entry, not in standings: its games count in no standings table, no league table, no Hype Cup input and no other setup's numbers
Entry date
2026-11-07first match 11:33 UTC
League field played
League 1 (original table)19 setups
Matches played
18 / 19GPT-6 LunaDecisions API: outside the requested scope (Ümit, 12 Nov); not played (opponent's API out of credit)
Record
11–1–6won–drawn–lost
Error rate (PR), 95% interval
27.824.9 to 30.6
Decisions
1,998 (1,998 scored)
Risk judgement, margin 0.001
60.558.7 to 62.4
Ability d′
0.930.78 to 1.06 · risks taken 249 / 666 Risk pays · needless 90 / 850 Safety pays
Caution c (+ cautious, − bold)
+0.78+0.70 to +0.87 · cautious
Engine's best play
51.7%
Median s per decision
1.74
$ per 1,000 decisions
1.545estimate · RTX PRO 6000 Blackwell
Invalid answers
0of 1,998 decisions
Cut off at the output limit
00 of 1,998 decisions sent an output limit; 0 over 32,768 output tokens without one
Route and effort sent
local (reax13 serve.py)reax/reax13 · effort not in the requests

A late entry joins an original league table after it was played and plays every setup of that table once, in matches of its own. Its games count in no standings table, no league table, no Hype Cup input and no other setup's numbers. The numbers here come from its own late-entry games only, computed as for a setup in a league.

Error rate and risk judgement: 95% intervals, whole games resampled. Ability d′ and caution c at the 0.001 margin: rates corrected by adding 0.5 to the counts and 1 to the totals; 95% intervals, blocks of two seat-swapped games resampled. Cut off: answers that stopped at an output limit the request sent (finish reason length, stop reason max_tokens, finishReason MAX_TOKENS, incomplete max_output_tokens), in the stored response or error. Effort: the reasoning effort the requests sent. Engine's best play: the share of its scored decisions that chose the engine's best play (equity loss under the best-play threshold). Time: median request seconds per decision (waits between retries not included). Cost per 1,000 decisions: the recorded tokens at the price shown, as labelled there (list or billed price; for a setup on our GPUs the hosted model's list price, not billed).

Late entries on the Leagues page · JSON

How we called it

Model ID
decisionmodels/REAX-1.3 @ fa4bf7ee3bb56a8bc421dad090631be1c5845c2d (Hugging Face, Apache-2.0; two parts in FP8 on one GPU)
Provider or endpoint
NVIDIA RTX PRO 6000 Blackwell (96 GB), in Docker, through reax-engine (github.com/reaxlabs/reax-engine @ ff8f595f149abdd50ea914455eadc74cf97a277f, reax13 mode, its serve.py), one request at a time, no outside network
Date run
7 Nov 2026
Price
no price per token; estimate: GPU time, at most; source content/own_cost.json (wall clock on our machines × hourly rate); price date not recorded
Model type
decision model
Weights
open
Runs on
our GPUs
Reasoning setting
none (decision model)
Confidence source
decision-model probability
Rules text
long
Option groups
up to 255 options per request (the server's limit, from the model card); more legal plays are split into equal groups, a winner per group, then among the winners; in this data 1 decisions were split into groups of 163
Sampling settings
not sent (decision request)
Notes
Late entry to the original League 1 table (not in standings, not eligible for the Hype Cup); its matches were played on 7 Nov 2026. Each option's probability is the 50/50 mean of the two parts' probabilities, stored as returned; in the smoke test before the run every probability sum was 1.000000. Its pair with GPT-6 Luna was not played: GPT-6 Luna's API was out of credit; the pair is outside the requested scope (Ümit, 12 Nov). Run note: the server's container ran without Docker's no-new-privileges setting, approved for this run only (the host's Docker refused every container with it); every other protection was kept: all capabilities dropped, read-only root filesystem, non-root user, weights mounted read-only, no outside network.

Matches

Swipe for more columns →

REAX 1.3 against each opponent, its score first: 11 won, 1 drawn, 6 lost; 1 outside the requested scope (Ümit, 12 Nov), not played, not a loss; none counts in a standings table
OpponentScoreResult#Finished
Random7–0Won9052026-11-07
GNUbg weak1–5Lost9062026-11-07
Safest6–0Won9072026-11-07
Cygnet2–5Lost9082026-11-07
Winnow-12B6–3Won9092026-11-07
Perplexity Decider Flash v1.25–0Won9102026-11-07
Perplexity Decider 27B v1.2NVIDIA RTX PRO 6000 Blackwell6–0Won9112026-11-07
GLiDEv1.2 · NVIDIA RTX PRO 6000 Blackwell6–2Won9122026-11-07
Clef6–0Won9132026-11-07
Open-Jev-27B-v1.1NVIDIA RTX PRO 6000 Blackwell5–1Won9142026-11-07
Perplexity Decider 27B v1.1NVIDIA RTX PRO 6000 Blackwell + NVIDIA DGX Spark1–5Lost9152026-11-07
celeris-1-decision4–5Lost9162026-11-07
GLiDEAPI5–5Draw9172026-11-07
Jev 1.135–0Won9182026-11-07
Jev 1.145–6Lost9192026-11-07
d1-ultra5–3Won9202026-11-07
Mercury Decide6–1Won9212026-11-07
SubQ Snap PreviewThe opponent was requested as SubQ Snap Preview; under that name the API returned answers identical to SubQ Snap 1.1 on 1,000 paired positions, so the opponent's model identity is not verified.2–6Lost9222026-11-07
GPT-6 LunaDecisions APINo scoreOutside the requested scope (Ümit, 12 Nov); not played (opponent's API out of credit)N/A