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

DeepSeek-V4.1-Flashthinking · first release

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
deepseek-v4.1-flash
Provider or endpoint
our GPUs (SGLang)
Date run
Missing
Price
no price per token; own hardware; price date not recorded
Model type
chat model, one letter
Weights
open
Runs on
our GPUs
Reasoning setting
thinking on, reasoning effort 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
answer constrained by a JSON schema; stored as choice only

Missing B7 fields: option_groups, 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)
18.115.6 to 20.8
Decisions
1,000 (1,000 scored, 0 invalid)
Engine's best play
61.8%
Blunder rate
14.4%
P95 equity loss
0.194
Mean loss, worst 5%
0.334
Median / P90 s per decision
105.8 / 287.5
$ per 1,000 decisions
Not measured
Risk judgement, margin 0.001
68.2 (95% interval 65.2 to 71.1); risks found 44.5% (158 / 355), right safety 91.9% (339 / 369)
Risk judgement, margin 0.02
69.0 (95% interval 66.0 to 72.2); risks found 47.2% (150 / 318), right safety 90.9% (369 / 406)
No trade-off
276 (left out of every risk metric)

JSON

Risk call card

68.2risk judgement, margin 0.001

right risks158
wrong risks27
needless risks30
missed risks170
right safety339

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

By game phase

PhaseError rate (PR)Decisions
OpeningToo few decisions143
Contact23.4641
Holding gameToo few decisions20
RaceToo few decisions136
Bear-offToo few decisions60