---
id: provider-cache-hit-rate/2026-08
series: provider-cache-hit-rate
title: Prompt cache hit rate by provider, August 2026
period: 2026-08
period_start: 2026-08-01
period_end: 2026-08-31
updated: 2026-11-11
license: CC BY 4.0
canonical: https://www.requesty.ai/data/provider-cache-hit-rate/2026-08
latest: https://www.requesty.ai/data/provider-cache-hit-rate
---

# Prompt cache hit rate by provider, August 2026

> DeepSeek served 94.7% of input tokens from cache in August 2026, the highest among major providers. The median major provider served 76.4%.

Question: What share of input tokens do LLM providers serve from prompt cache?

## Key findings

1. DeepSeek leads at 94.7%; Alibaba is lowest at 42.7%.
2. 6 of 17 major providers served at least 80% of input tokens from cache.
3. Novita moved the most since July 2026: cache hit rate went from 38.7% to 93.2% (+54.5 pp).

## Data

| Provider | Cache hit rate | Traffic share | Organizations (band) |
| --- | --- | --- | --- |
| DeepSeek | 94.67% | 8.75% | 500+ |
| Novita | 93.22% | 1.63% | 500+ |
| Fireworks | 90.78% | 2.95% | 500+ |
| Sference | 88.85% | 4.63% | 500+ |
| Anthropic | 86.46% | 8.16% | 500+ |
| DeepInfra | 80.99% | 5.92% | 500+ |
| Vertex (Claude) | 79.85% | 6.96% | 500+ |
| Bedrock | 78.86% | 3.85% | 500+ |
| Azure | 76.39% | 8.33% | 500+ |
| Nebius | 75.89% | 1.00% | 500+ |
| NVIDIA | 73.46% | 1.00% | 500+ |
| xAI | 67.41% | 0.79% | 100-499 |
| TensorX | 61.74% | 4.44% | 500+ |
| Vertex (Gemini) | 52.78% | 20.53% | 500+ |
| OpenAI | 50.28% | 17.24% | 500+ |
| Mistral | 48.95% | 1.52% | 500+ |
| Alibaba | 42.72% | 0.66% | 100-499 |

## Method

- Cache hit rate is cached input tokens divided by input tokens, the same definition /rankings uses.
- Providers that report no cached tokens are left out of the chart. Only providers with at least 0.5% of published traffic are shown.
- A row is published only when it has at least 20 requests, at least 5 distinct organizations, and no single organization above 80% of its traffic. These are the same floors /rankings uses. Rows below the floors are withheld, not estimated.
- Only shares, rates, medians and organization bands are published. Absolute request, token and spend totals stay private.

## Caveats

- Hit rate depends on workload. Long agent sessions with a stable prefix cache far better than one-off requests.
- Providers report cached tokens differently. Some only count explicit cache reads.

## Cite as

Requesty (2026). Prompt cache hit rate by provider, August 2026. Requesty Data. https://www.requesty.ai/data/provider-cache-hit-rate/2026-08

```bibtex
@misc{requesty_provider_cache_hit_rate_2026_08,
  author       = {{Requesty}},
  title        = {Prompt cache hit rate by provider, August 2026},
  year         = {2026},
  howpublished = {\url{https://www.requesty.ai/data/provider-cache-hit-rate/2026-08}},
  note         = {Requesty Data}
}
```

Source: Requesty production gateway. License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/).
Downloads: [JSON](https://www.requesty.ai/data/provider-cache-hit-rate/2026-08/data.json) · [CSV](https://www.requesty.ai/data/provider-cache-hit-rate/2026-08/data.csv) · [Markdown](https://www.requesty.ai/data/provider-cache-hit-rate/2026-08/data.md)