TIS-100 Bench: T31 Node Emulator

Emulate eight addressable memory slots using read and write commands. Stack memory must be organized to support indexed access.

In TIS-100 Bench, 16 LLMs each had one attempt at this TIS-100 puzzle, with 45 minutes to find a working program and keep improving it, without seeing the community leaderboard. 15 of 16 models solved it. Together they found 4 new Pareto-frontier points beyond the community leaderboard.

15 / 16models solved it
215best LLM cycles (151 community)
4new frontier points

Records

ScoreCommunity bestBest LLMModel
Cycles151215Opus 5.5 · +42%
Nodes34DeepSeek V4.1 Flash · +33%
Instructions2328GPT-6 Astra · +22%

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
01Opus 5.5Yes2154283
02Fable 5.1Yes3234291
03Muse Wonder 1.0Yes327435—
04Sonnet 5.5Yes352429—
05Gemini 3.8 FlashYes358429—
06GPT-6 AstraYes365428—
07GPT-6.1 SolYes387428—
08Haiku 5.5Yes405431—
09MiMo-V2.6-ProYes451432—
10Gemini 4 ArgonYes481428—
11Kimi K3.1 ProYes531428—
12DeepSeek V4.1 FlashYes639438—
13GPT-6 LunaYes641431—
14GLM-5.3Yes734433—
15Grok 4.7Yes1,077440—
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.

Opus 5.5 Fewest cycles★ New frontier point

215 cycles · 6 nodes · 63 instructionsNEXUS.23.727.9.215-6-63.txt on GitHub

Opus 5.5 program running on NEXUS.23.727.9 in TIS-100
Node 0
S: MOV UP, ACC
JEZ W
MOV UP, ACC
MOV ACC, RIGHT
MOV ACC, RIGHT
MOV 0, RIGHT
JMP S
W: MOV UP, ACC
MOV ACC, RIGHT
MOV ACC, RIGHT
MOV UP, RIGHT
Node 1
SUB LEFT
MOV ACC, DOWN
MOV LEFT, ACC
MOV LEFT, DOWN
Node 3
S: JRO RIGHT
MOV UP, DOWN
MOV UP, DOWN
MOV UP, DOWN
MOV UP, DOWN
JMP O
MOV DOWN, UP
MOV DOWN, UP
MOV DOWN, UP
MOV DOWN, UP
MOV DOWN, UP
MOV DOWN, UP
MOV DOWN, UP
O: MOV DOWN, RIGHT
MOV RIGHT, DOWN
Node 4
S: MOV RIGHT, LEFT
MOV RIGHT, ACC
JGZ W
JLZ E
MOV LEFT, ACC
MOV ACC, LEFT
MOV ACC, DOWN
JMP S
E: MOV LEFT, LEFT
JMP S
W: MOV LEFT, NIL
MOV ACC, LEFT
Node 5
S: MOV UP, ACC
JGZ B
ADD 13
JMP E
B: NEG
ADD 5
JGZ E
MOV 1, LEFT
MOV -2, LEFT
ADD 4
E: MOV ACC, LEFT
MOV UP, LEFT
Node 7
MOV 2, ACC
I: MOV 0, LEFT
MOV 0, LEFT
MOV 0, LEFT
MOV 0, LEFT
SUB 1
JNZ I
S: MOV UP, DOWN
JMP S

Fable 5.1 Fewest nodes★ New frontier point

323 cycles · 4 nodes · 43 instructionsNEXUS.23.727.9.323-4-43.txt on GitHub

Fable 5.1 program running on NEXUS.23.727.9 in TIS-100
Node 0
L: MOV UP, ACC
SAV
MOV UP, ACC
SUB 8
MOV ACC, DOWN
SWP
MOV ACC, DOWN
JNZ X
MOV UP, DOWN
X: SWP
MOV ACC, DOWN
Node 3
S: SUB RIGHT
JGZ P
JEZ D
N: MOV UP, DOWN
ADD 1
JLZ N
JMP D
P: MOV DOWN, UP
SUB 1
JGZ P
D: MOV RIGHT, ACC
MOV UP, RIGHT
MOV RIGHT, UP
Node 4
L: MOV UP, LEFT
MOV UP, ACC
JEZ W
MOV UP, LEFT
MOV LEFT, ACC
MOV ACC, DOWN
MOV ACC, LEFT
JMP L
W: MOV UP, ACC
MOV UP, LEFT
MOV LEFT, NIL
MOV ACC, LEFT
Node 7
SUB 4
I: MOV 0, LEFT
MOV 0, LEFT
ADD 1
JLZ I
L: MOV UP, DOWN
JMP L

GPT-6 Astra Fewest instructions

748 cycles · 4 nodes · 28 instructionsNEXUS.23.727.9.748-4-28.txt on GitHub

GPT-6 Astra program running on NEXUS.23.727.9 in TIS-100
Node 0
MOV 9 ACC
I: MOV 0 LEFT
SUB 1
JNZ I
R: MOV UP DOWN
JMP R
Node 3
R: MOV 8 ACC
SUB RIGHT
SAV
L: MOV UP DOWN
SUB 1
JGZ L
A: MOV UP RIGHT
MOV RIGHT UP
SWP
B: MOV DOWN UP
SUB 1
JGZ B
Node 4
R: MOV UP ACC
MOV UP LEFT
JEZ W
MOV LEFT ACC
MOV ACC LEFT
MOV ACC DOWN
JMP R
W: MOV LEFT NIL
MOV UP LEFT
Node 7
MOV UP DOWN