TIS-100 Bench: Error: 8

A garbled specification describes a counter that advances by 8 per input pair. Emit the upper bound if the counter is above it, the lower bound if below it, and zero otherwise.

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. 11 of 16 models solved it. Together they found 12 new Pareto-frontier points beyond the community leaderboard.

11 / 16models solved it
186best LLM cycles (198 community)
12new frontier points

Records

ScoreCommunity bestBest LLMModel
Cycles198186Gemini 4 Argon ★ New record
Nodes44Fable 5.1 Ties record
Instructions1720Fable 5.1 · +18%

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
01Fable 5.1Yes2254205
02Gemini 4 ArgonYes1864203
03GPT-6 AstraYes2234212
04Sonnet 5.5Yes2884212
05Opus 5.5Yes2034221
06GPT-6.1 SolYes189420—
07Kimi K3.1 ProYes216423—
08Gemini 3.8 FlashYes280422—
09Muse Wonder 1.0Yes308421—
10GLM-5.3Yes354425—
11Grok 4.7Yes460746—
12DeepSeek V4.1 FlashNo————
13Haiku 5.5No————
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.

Gemini 4 Argon Fewest cycles

186 cycles · 9 nodes · 54 instructionsSPEC79324720.186-9-54.txt on GitHub

Gemini 4 Argon program running on SPEC79324720 in TIS-100
Node 0
ADD 8
MOV ACC, RIGHT
ADD 8
Node 1
MOV UP, DOWN
MOV LEFT, ACC
SAV
SUB UP
JLZ P
SWP
P: MOV ACC, RIGHT
Node 2
MOV UP, DOWN
MOV UP, RIGHT
MOV LEFT, RIGHT
Node 3
ADD 8
SAV
SUB LEFT
JLZ K
MOV ACC, DOWN
MOV LEFT, ACC
JMP E
K: SWP
MOV LEFT, DOWN
E: ADD 8
Node 4
SAV
SUB UP
JLZ P
SWP
MOV ACC, RIGHT
JMP E
P: MOV ACC, RIGHT
SWP
E: ADD 16
Node 5
SAV
SUB UP
JLZ K
MOV ACC, DOWN
MOV LEFT, ACC
JMP E
K: SWP
MOV LEFT, DOWN
E: ADD 16
Node 6
ADD 8
SAV
SUB UP
MOV ACC, DOWN
SWP
ADD 8
Node 7
SAV
SUB UP
MOV ACC, DOWN
SWP
ADD 16
MOV RIGHT, DOWN
Node 8
MOV UP, LEFT

Sonnet 5.5 Fewest nodes★ New frontier point

303 cycles · 4 nodes · 21 instructionsSPEC79324720.303-4-21.txt on GitHub

Sonnet 5.5 program running on SPEC79324720 in TIS-100
Node 1
S: ADD 8
SAV
SUB UP
JGZ NN
MOV ACC, RIGHT
SWP
JMP S
NN: SWP
MOV ACC, RIGHT
Node 2
MOV LEFT, ACC
SAV
SUB UP
JGZ O
SWP
O: MOV ACC, DOWN
Node 5
MOV UP, DOWN
Node 7
ADD 8
SAV
SUB UP
MOV ACC, DOWN
SWP

Fable 5.1 Fewest instructions

375 cycles · 5 nodes · 20 instructionsSPEC79324720.375-5-20.txt on GitHub

Fable 5.1 program running on SPEC79324720 in TIS-100
Node 1
MOV UP, RIGHT
Node 2
L: MOV LEFT, ACC
ADD RIGHT
JGZ N
MOV UP, ACC
SAV
ADD RIGHT
JLZ M
MOV 0, DOWN
JMP L
N: MOV UP, NIL
SUB RIGHT
JMP O
M: SWP
O: MOV ACC, DOWN
Node 3
SUB 8
MOV ACC, LEFT
MOV ACC, LEFT
Node 5
MOV UP, DOWN
Node 7
MOV UP, DOWN