TIS-100 Bench: Signal Comparator
Classify each input as positive, zero, or negative. Every input produces three outputs: one 1 and two 0s.
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. 15 of 16 models solved it. No model found a new Pareto-frontier point beyond the community leaderboard.
Records
| Score | Community best | Best LLM | Model |
|---|---|---|---|
| Cycles | 188 | 227 | GPT-6 Astra · +21% |
| Nodes | 6 | 6 | DeepSeek V4.1 Flash Ties record |
| Instructions | 15 | 17 | Gemini 4 Argon · +13% |
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 ↙Search over time
Loading puzzle trajectories…
The search over time
Best in-game score so far, before independent testing. Each score can come from a different program, and lines stop when the episode ends.
Every model
| Model | Solved | Best cycles | Best nodes | Best instructions | New frontier points |
|---|---|---|---|---|---|
| 01GPT-6 Astra | Yes | 227 | 6 | 18 | — |
| 02GPT-6.1 Sol | Yes | 228 | 6 | 20 | — |
| 03Fable 5.1 | Yes | 245 | 6 | 18 | — |
| 04Gemini 4 Argon | Yes | 245 | 6 | 17 | — |
| 05Opus 5.5 | Yes | 252 | 6 | 18 | — |
| 06DeepSeek V4.1 Flash | Yes | 265 | 6 | 20 | — |
| 07GLM-5.3 | Yes | 265 | 6 | 20 | — |
| 08Gemini 3.8 Flash | Yes | 265 | 6 | 20 | — |
| 09Muse Spark 1.3 | Yes | 265 | 6 | 19 | — |
| 10Muse Wonder 1.0 | Yes | 265 | 6 | 20 | — |
| 11Sonnet 5.5 | Yes | 265 | 6 | 20 | — |
| 12GPT-6 Luna | Yes | 265 | 6 | 20 | — |
| 13Kimi K3.1 Pro | Yes | 266 | 6 | 20 | — |
| 14Grok 4.7 | Yes | 278 | 6 | 20 | — |
| 15Haiku 5.5 | Yes | 278 | 6 | 20 | — |
| 16MiMo-V2.6-Pro | No | — | — | — | — |
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
227 cycles · 6 nodes · 27 instructions21340.227-6-27.txt on GitHub

S: MOV UP ACC
JEZ A
JLZ B
MOV 1 DOWN
JMP S
A: MOV 4 DOWN
JMP S
B: MOV 6 DOWNMOV UP DOWNMOV UP RIGHTS: JRO LEFT
MOV 4 RIGHT
MOV 1 DOWN
JMP S
MOV 1 RIGHT
JMP C
MOV 6 RIGHT
C: MOV 0 DOWNS: JRO LEFT
MOV 0 RIGHT
MOV 1 DOWN
JMP S
MOV 0 RIGHT
JMP C
MOV 1 RIGHT
C: MOV 0 DOWNMOV LEFT DOWNGemini 4 Argon
398 cycles · 6 nodes · 17 instructions21340.398-6-17.txt on GitHub

MOV UP, DOWNMOV UP, DOWNMOV UP, RIGHTMOV LEFT, ACC
JGZ G
MOV 0, DOWN
JEZ Z
MOV 1, RIGHT
JMP Z
G: MOV 1, DOWN
MOV 0, RIGHT
Z: MOV 0, RIGHT
JNZ D
MOV 1, RIGHT
D:MOV LEFT, RIGHT
MOV LEFT, DOWNMOV LEFT, DOWN