# How Vercel uses Fireworks AI

**40X**: “a 40X improvement in end-to-end latency for v0 users”

As published on [fireworks.ai](https://fireworks.ai/customers/vercel-v0-code-fixing-case-study), November 2025. Captured by usedby on 2026-10-31.

- Company: [Vercel](https://www.usedby.ai/companies/vercel.md)
- Tool: [Fireworks AI](https://www.usedby.ai/tools/fireworks-ai.md)
- Industry: [DevTools & Infrastructure](https://www.usedby.ai/companies/industry/devtools-infrastructure.md)

## The use case: how Vercel deploys Fireworks AI

Vercel uses Fireworks AI to run and fine-tune an open-source model that automatically fixes errors in code generated by its v0 tool. It applies reinforcement fine-tuning and speculative decoding to improve error-free output and speed.

Summary written by usedby from the source page, in English. The figures are those of Fireworks AI and Vercel, not ours.

> The result is a platform capable of achieving a 93% error-free generation rate and a 40X improvement in end-to-end latency for v0 users, setting a new benchmark for developer-facing AI tooling.

- **93%**: “The result is a platform capable of achieving a 93% error-free generation rate and a 40X improvement in end-to-end latency for v0 users, setting a new benchmark for developer-facing AI tooling.”
- **62%**: “In our evaluation, Sonnet 3.5 compiled at 62%, and we got our error-free generation rate well into the 90s”

> Vercel's v0 model is a composite model. The SOTA in this space changes every day, so you don't want to tie yourself to a single model. Using a fine-tuned reinforcement learning model with Fireworks, we perform substantially better than SOTA. In our evaluation, Sonnet 3.5 compiled at 62%, and we got our error-free generation rate well into the 90s
>
> Malte Ubl, CTO at Vercel

Source: [fireworks.ai](https://fireworks.ai/customers/vercel-v0-code-fixing-case-study), captured 2026-10-31.

> Vercel’s v0 model is a composite model. The SOTA in this space changes every day, so you don’t want to tie yourself to a single model. Using a fine-tuned reinforcement learning model with Fireworks, we perform substantially better than SOTA.
>
> Malte Ubl, CTO

Source: [fireworks.ai](https://fireworks.ai/), captured 2026-10-05.

## Other AI tools used by Vercel

- [AWS Bedrock](https://www.usedby.ai/tools/aws-bedrock.md) — Confirmed
- [Notion](https://www.usedby.ai/tools/notion.md) — Confirmed
- [Baseten](https://www.usedby.ai/tools/baseten.md) — Observed
- [Braintrust](https://www.usedby.ai/tools/braintrust.md) — Observed
- [BrightHire](https://www.usedby.ai/tools/brighthire.md) — Observed
- [Browserbase](https://www.usedby.ai/tools/browserbase.md) — Observed

## Fireworks AI users in DevTools & Infrastructure

- [Cursor](https://www.usedby.ai/companies/cursor.md) — Confirmed
- [Hasura](https://www.usedby.ai/companies/hasura.md) — Declared by the company
- [Linear](https://www.usedby.ai/companies/linear.md) — Declared by the company
- [Replit](https://www.usedby.ai/companies/replit.md) — Declared by the company
- [Turbopuffer](https://www.usedby.ai/companies/turbopuffer.md) — Declared by the company
- [Cognition](https://www.usedby.ai/companies/cognition.md) — Observed

## What usedby checked

We compared the story with its live page on 2026-10-31.

- Checked: the figure 40X is printed word for word on the page, near the name of Vercel.
- Checked: the passage quoted above is copied word for word from the page, near the name of Vercel.
- Checked: each number in our summary is printed on the page.
- Checked: the publication date is read from the page’s own metadata, never guessed.
- Checked: Vercel uses Fireworks AI. Confirmed line. Latest check across sources: 2026-10-31.
- Not checked: the result itself. We quote it; we did not measure it.

---
Source: https://www.usedby.ai/case-studies/vercel-fireworks-ai · How we check: https://www.usedby.ai/methodology