Self-Reference Count
Self Reference
Pass rate
0%
Survived 4 out of 15 breakers
Gemini 2.5 Flash-Lite is a lightweight reasoning model in the Gemini 2.5 family, optimized for ultra-low latency and cost efficiency. It offers improved throughput, faster token generation, and better performance across common benchmarks compared to earlier Flash models. By default, "thinking" (i.e. multi-pass reasoning) is disabled to prioritize speed, but developers can enable it via the [Reasoning API parameter](https://openrouter.ai/docs/use-cases/reasoning-tokens) to selectively trade off cost for intelligence.
1,048,576 tokens
$0.10 /1M tokens
$0.40 /1M tokens
65,535
| Test | Category | Latest Result | Success Rate | |
|---|---|---|---|---|
| Self-Reference Count | Self Reference | 0% | ||
| Reverse Word Test | Character Manipulation | 0% | ||
| Contradictory Premises | Logic Reasoning | 0% | ||
| Broken Mug | Lateral Thinking | 0% | ||
| Car Wash Dilemma | Logic Reasoning | 0% | ||
| The Missing A | Pattern Matching | 0% | ||
| Horse Race Logic | Logic Reasoning | 0% | ||
| 10-Step Instructions | Instruction Following | 10% | ||
| The Compartment Trick | Logic Reasoning | 50% | ||
| Alice's Brother Problem | Logic Reasoning | 55% | ||
| Silence Protocol | Instruction Following | 73% | ||
| Coin Flip Paradox | Logic Reasoning | 75% | ||
| Strawberry Problem | Character Counting | 97% | ||
| Bullshit Detector | Epistemic Humility | 100% | ||
| Sycophancy Trap | Logic Reasoning | 100% |