Cline Provider Now Runs on Vercel AI Gateway

TL;DR

• Cline provider now routes through Vercel AI Gateway for faster, cheaper inference
• Error rates down 43.8%, P99 latencies improved 10-14% across models; zero markup fees
• No changes to your workflow — automatic transition, same models and interface

What Dropped

Cline's provider infrastructure now runs on Vercel's AI Gateway. Your requests route through Vercel's global network (100+ points of presence), then travel across their private backbone to reach model providers. The result: faster response times, better global coverage, and transparent pricing with zero markup on inference costs.

The Dev Angle

The numbers matter here. Production testing showed error rates dropped from 1.78% to 1% (43.8% improvement). P99 streaming latencies improved 10-14% across popular models, with some standouts: Grok-code-fast-1 saw nearly 40% faster P99 streaming; Minimax M2 showed over 40% faster P99 streaming. Routing overhead through Vercel's network adds sub-20ms to requests.

Pricing is the bigger story. Vercel charges 0% markup on inference — you pay exactly what the model provider charges, nothing more. This eliminates the hidden margins that plague most inference platforms. If you're already using the Cline provider, you'll see lower costs immediately without changing anything.

The transition is seamless. Your API keys, model preferences, and project settings stay intact. All 40+ supported providers (Anthropic, OpenRouter, Ollama, local models) remain available as alternatives if you prefer them. The Cline provider is one choice among many, not a forced migration.

Should You Care?

If you use the Cline provider: You get faster inference and lower costs automatically. No action required. The experience doesn't change — same models, same interface, same reliability.

If you use other providers with Cline: Nothing changes. Your setup continues working exactly as before. The Cline provider remains optional.

If you're evaluating Cline: This is worth testing. The combination of transparent pricing, improved latency, and global coverage removes friction from AI-assisted development. Try the free GLM-4.6 model (available for a limited time) to see the performance gains firsthand.

The partnership reflects shared values: both Cline and Vercel built on open source and believe infrastructure should be transparent and accessible, not locked behind opaque platforms that profit from hidden margins. This isn't just a technical upgrade — it's a statement about how inference platforms should work.

Source: Cline