chat model, one letter · v1 set (earlier version)

Gemini 4 Argon

Earlier version

Results f24d1f41, updated 16 November 2026, 22:20 UTC; scored by GNU Backgammon 1.08, 2-ply.

How we called it

Model ID
gemini-4-argon
Provider or endpoint
Gemini API
Date run
Missing
Price
$2.0 / $10.0 per million tokens in / out; API price; price date not recorded
Model type
chat model, one letter
Weights
closed
Runs on
API
Reasoning setting
thinking level low
Confidence source
none
Rules text
short (the v1 rules text; the same stored request text for every model)
Option groups
not recorded; no decision in this data was split
Sampling settings
temperature 1.0
Notes
no token probabilities; answer constrained to the option letters, stored as choice only

Missing B7 fields: option_groups, price.date, date_run

Results in v1 set (earlier version)

Earlier version: plays were listed in the runner's order, which tends to put the best play early. When Gemini 4 Argon answered the same 1,000 positions with the plays shuffled, its error rate changed by +1.9 (95% interval −0.2 to +4.2), so the effect is not proven.

Error rate (PR)
16.914.7 to 19.3
Decisions
1,000 (1,000 scored, 0 invalid)
Engine's best play
62.8%
Blunder rate
13.7%
P95 equity loss
0.189
Mean loss, worst 5%
0.312
Median / P90 s per decision
4.08 / 28.0
$ per 1,000 decisions
Not measured
Risk judgement, margin 0.001
70.2 (95% interval 67.1 to 73.1); risks found 50.4% (179 / 355), right safety 90.0% (332 / 369)
Risk judgement, margin 0.02
69.8 (95% interval 66.8 to 73.0); risks found 51.9% (165 / 318), right safety 87.7% (356 / 406)
No trade-off
276 (left out of every risk metric)

JSON

Risk call card

70.2risk judgement, margin 0.001

right risks179
wrong risks30
needless risks37
missed risks146
right safety332

Hit risk only: being hit. Gammon risk and priming are not in it.

By game phase

PhaseError rate (PR)Decisions
OpeningToo few decisions143
Contact22.4641
Holding gameToo few decisions20
RaceToo few decisions136
Bear-offToo few decisions60