# Test-Time Compute

> Spending more compute during inference to improve answer quality without retraining the model.

- Category: Reasoning & Inference
- Stage: Emerging
- Also called: Inference-time scaling, Extra thinking
- Canonical: https://www.joelcsaji.com/ai-dictionary/test-time-compute

## Plain English

The model gets more time or attempts to think when the task is hard.

## Example

A coding benchmark run gives an agent multiple attempts, tool calls, or verifier passes before selecting a final patch.

## Why it matters

The frontier is moving from only bigger training runs to smarter inference strategies.

## Related

- [Reasoning Model](https://www.joelcsaji.com/ai-dictionary/reasoning-model)
- [Parallelization](https://www.joelcsaji.com/ai-dictionary/parallelization)
- [Judge Model](https://www.joelcsaji.com/ai-dictionary/judge-model)

## Sources

- [Stanford HAI: 2026 AI Index](https://hai.stanford.edu/ai-index/2026-ai-index-report)
- [Anthropic: Building Effective Agents](https://www.anthropic.com/engineering/building-effective-agents)