TIS-100 Bench: Tri-Sequence Identifier

Identify which of three repeating numeric sequences a subsequence belongs to, then predict its next value.

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. 7 of 16 models solved it. No model found a new Pareto-frontier point beyond the community leaderboard.

7 / 16models solved it
192best LLM cycles (177 community)
0new frontier points

Records

ScoreCommunity bestBest LLMModel
Cycles177192GPT-6 Astra · +8%
Nodes34GPT-6 Astra · +33%
Instructions2940Opus 5.5 · +38%

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
01GPT-6 AstraYes192447—
02GPT-6.1 SolYes197444—
03Gemini 3.8 FlashYes225558—
04Muse Wonder 1.0Yes244446—
05Gemini 4 ArgonYes246447—
06Sonnet 5.5Yes290451—
07Opus 5.5Yes317440—
08DeepSeek V4.1 FlashNo————
09Fable 5.1No————
10GLM-5.3No————
11Grok 4.7No————
12Haiku 5.5No————
13Kimi K3.1 ProNo————
14MiMo-V2.6-ProNo————
15Muse Spark 1.3No————
16GPT-6 LunaNo————

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.

GPT-6 Astra Fewest cycles

192 cycles · 5 nodes · 49 instructionsSPEC26159035.192-5-49.txt on GitHub

GPT-6 Astra program running on SPEC26159035 in TIS-100
Node 0
L:MOV RIGHT ACC
JEZ E
SAV
JMP L
E:SWP
SUB RIGHT
ADD 4
MOV ACC DOWN
Node 1
MOV UP ACC
SWP
MOV UP ACC
L:SWP
MOV ACC LEFT
MOV UP ACC
JNZ L
MOV 0 LEFT
SWP
MOV ACC LEFT
Node 3
L:MOV RIGHT ACC
SUB 67
JEZ FOUR
JGZ EIGHT
ADD 67
JRO ACC
JMP A
JMP EIGHT
FOUR:MOV 4 RIGHT
JMP L
EIGHT:MOV 8 RIGHT
JMP L
A:MOV 16 RIGHT
JMP L
MOV 2 RIGHT
Node 4
L:MOV UP ACC
JGZ P
JEZ ONE
ADD 44
JGZ S
ADD 8
JEZ B
ONE:MOV 1 DOWN
JMP L
S:MOV 64 DOWN
JMP L
B:MOV 512 DOWN
JMP L
P:MOV ACC LEFT
MOV LEFT DOWN
Node 7
MOV UP DOWN

GPT-6.1 Sol Fewest nodes

197 cycles · 4 nodes · 49 instructionsSPEC26159035.197-4-49.txt on GitHub

GPT-6.1 Sol program running on SPEC26159035 in TIS-100
Node 0
L:MOV RIGHT ACC
JEZ E
SAV
JMP L
E:SWP
SUB RIGHT
JLZ P
SUB 7
JEZ P
SUB 56
JEZ B
MOV -1 ACC
B:SUB 1
P:MOV ACC DOWN
Node 1
MOV UP ACC
SAV
MOV UP ACC
L:SWP
MOV ACC LEFT
MOV UP ACC
JNZ L
MOV ACC LEFT
SWP
MOV ACC LEFT
Node 4
L:MOV UP ACC
ADD 4
JGZ P
JEZ O
ADD 8
JLZ H
MOV 7 DOWN
JMP L
H:ADD 44
JEZ F
O:MOV 5 DOWN
JMP L
F:ADD 9
P:MOV ACC DOWN
Node 7
L:MOV 1 ACC
JRO UP
ADD ACC
ADD ACC
ADD ACC
ADD ACC
MOV ACC DOWN
JMP L
MOV 64 DOWN
JMP L
MOV 512 DOWN

Opus 5.5 Fewest instructions

391 cycles · 4 nodes · 40 instructionsSPEC26159035.391-4-40.txt on GitHub

Opus 5.5 program running on SPEC26159035 in TIS-100
Node 0
MOV RIGHT, ACC
ADD -505
JGZ S
ADD 449
JGZ S
ADD 51
JGZ S
ADD 13
JGZ S
ADD 5
JGZ S
ADD 39
JGZ S
S: MOV ACC, DOWN
Node 1
MOV ACC, DOWN
L: MOV UP, ACC
JEZ Z
MOV ACC, DOWN
SWP
SUB DOWN
MOV DOWN, NIL
MOV ACC, DOWN
JMP L
Z: SWP
MOV DOWN, LEFT
Node 4
MOV UP, ACC
JGZ J
ADD 7
JGZ J
ADD 393
J: JRO ACC
ADD 63
ADD -509
ADD 512
ADD -14
ADD 9
ADD 5
ADD -3
MOV ACC, DOWN
Node 7
MOV UP, DOWN