TIS-100 Bench: Signal Averager

Average two inputs and round down. Neither averaging nor division is a native instruction.

In TIS-100 Bench, 16 LLMs each had one attempt at this TIS-100 puzzle, with 10 minutes to find a working program and keep improving it, without seeing the community leaderboard. 12 of 16 models solved it. No model found a new Pareto-frontier point beyond the community leaderboard.

12 / 16models solved it
1,181best LLM cycles (432 community)
0new frontier points

Records

ScoreCommunity bestBest LLMModel
Cycles4321,181Gemini 4 Argon · +173%
Nodes44Fable 5.1 Ties record
Instructions1023Fable 5.1 · +130%

Community records from the community leaderboard when we froze it. A record counts only after fresh-game replay and 100,000 passing random tests.

Pareto frontier

Pareto explorer

Lower is better ↙
Loading interactive Pareto explorer…
Community leaderboardModel frontier pointNew frontier point
Select a point with the mouse or keyboard to inspect its scores and program.

Search over time

Loading puzzle trajectories…

Every model

ModelSolvedBest cyclesBest nodesBest instructionsNew frontier points
01Gemini 4 ArgonYes1,181441—
02Sonnet 5.5Yes1,639428—
03GPT-6 AstraYes1,670424—
04Opus 5.5Yes1,937434—
05Muse Wonder 1.0Yes1,943567—
06Grok 4.7Yes2,030894—
07GPT-6.1 SolYes2,289423—
08Fable 5.1Yes2,327423—
09Gemini 3.8 FlashYes2,601428—
10GPT-6 LunaYes5,524546—
11Kimi K3.1 ProYes29,818434—
12Haiku 5.5Yes30,705430—
13DeepSeek V4.1 FlashNo————
14GLM-5.3No————
15MiMo-V2.6-ProNo————
16Muse Spark 1.3No————

Best scores can come from different programs. New frontier points are not matched or beaten on all three scores by any community program.

Best model programs

These programs are in TIS-100 save format and released under the MIT License. All of this puzzle’s model programs are on GitHub and Hugging Face.

Gemini 4 Argon Fewest cycles

1,181 cycles · 6 nodes · 81 instructionsNEXUS.06.786.0.1181-6-81.txt on GitHub

Gemini 4 Argon program running on NEXUS.06.786.0 in TIS-100
Node 1
MOV 0, ACC
SAV
MOV UP, ACC
L: SUB 256
JLZ D
SWP
ADD 128
SWP
JMP L
D: MOV ACC, RIGHT
SWP
MOV ACC, RIGHT
Node 2
MOV LEFT, ACC
ADD UP
SAV
MOV LEFT, ACC
SWP
L: JLZ D
SWP
ADD 128
SWP
SUB 256
JMP L
D: MOV ACC, DOWN
SWP
MOV ACC, DOWN
Node 5
MOV UP, ACC
ADD 192
SAV
MOV UP, ACC
SWP
L: JLZ D
SWP
ADD 32
SWP
SUB 64
JMP L
D: MOV ACC, RIGHT
SWP
MOV ACC, RIGHT
Node 6
MOV LEFT, ACC
ADD 48
SAV
MOV LEFT, ACC
SWP
L: JLZ D
SWP
ADD 8
SWP
SUB 16
JMP L
D: MOV ACC, DOWN
SWP
MOV ACC, DOWN
Node 9
MOV RIGHT, ACC
ADD 2
SAV
MOV RIGHT, ACC
SWP
L: JLZ D
SWP
ADD 1
SWP
SUB 2
JMP L
D: SWP
MOV ACC, DOWN
Node 10
MOV UP, ACC
ADD 12
SAV
MOV UP, ACC
SWP
L: JLZ D
SWP
ADD 2
SWP
SUB 4
JMP L
D: MOV ACC, LEFT
SWP
MOV ACC, LEFT

Gemini 4 Argon Fewest nodes

1,865 cycles · 4 nodes · 53 instructionsNEXUS.06.786.0.1865-4-53.txt on GitHub

Gemini 4 Argon program running on NEXUS.06.786.0 in TIS-100
Node 1
MOV 0, ACC
SAV
MOV UP, ACC
L: SUB 160
JLZ D
SWP
ADD 80
SWP
JMP L
D: MOV ACC, RIGHT
SWP
MOV ACC, RIGHT
Node 2
MOV LEFT, ACC
ADD UP
SAV
MOV LEFT, ACC
SWP
L: JLZ D
SWP
ADD 80
SWP
SUB 160
JMP L
D: MOV ACC, DOWN
SWP
MOV ACC, DOWN
Node 5
MOV UP, ACC
ADD 142
SAV
MOV UP, ACC
SWP
L: JLZ D
SWP
ADD 9
SWP
SUB 18
JMP L
D: MOV ACC, DOWN
SWP
MOV ACC, DOWN
Node 9
MOV UP, ACC
ADD 16
SAV
MOV UP, ACC
SWP
L: JLZ D
SWP
ADD 1
SWP
SUB 2
JMP L
D: SWP
MOV ACC, DOWN

GPT-6.1 Sol Fewest instructions

28,956 cycles · 4 nodes · 23 instructionsNEXUS.06.786.0.28956-4-23.txt on GitHub

GPT-6.1 Sol program running on NEXUS.06.786.0 in TIS-100
Node 1
MOV UP RIGHT
Node 2
MOV 999 ACC
SUB UP
SUB LEFT
MOV ACC DOWN
Node 5
MOV UP ACC
JLZ N
P: SUB 2
JLZ E
MOV 3 DOWN
JMP P
N: MOV 1 DOWN
ADD 2
JLZ N
E: MOV 5 DOWN
Node 9
S: MOV 499 ACC
L: JRO UP
ADD 1
JMP L
SUB 1
JMP L
MOV ACC DOWN
JMP S