Netlify added Claude Fable 5.1 and Gemini 3.8 Flash to AI Gateway and Agent Runners, but independent comparisons still favor Vercel's broader gateway.
Netlify added Anthropic’s Claude Fable 5.1 and Google’s Gemini 3.8 Flash to its AI Gateway and Agent Runners this week, both reachable with zero setup from any Netlify Function or agent run (Netlify changelog, Sept 1, 2026; Netlify changelog, Sept 2, 2026).
What changed #
Both models slot into the two AI products Netlify built on top of its credit-based plans: AI Gateway, a proxy that lets Functions call a model without holding a provider API key, and Agent Runners, which runs Claude Code, Codex or Gemini against a live Netlify project from the dashboard. A Function can now call Claude Fable 5.1 with the Anthropic SDK exactly as if it held a real Anthropic key, and Netlify handles the credentials, caching, rate limiting and auth underneath it (Netlify changelog, Sept 1, 2026). Gemini 3.8 Flash reaches both surfaces the same way a day later (Netlify changelog, Sept 2, 2026).
Nothing about the billing model changes. AI Gateway and Agent Runners draw from the same credit pool as builds and bandwidth, with model spend converted at a fixed 180 credits per $1 of underlying provider cost, and a team can cap that spend per member to stop one runaway agent run from burning the shared allotment (Netlify docs, pricing for AI features).
Why it matters #
This is Netlify catching its model list up, not closing a product gap. Vercel’s AI Gateway, by contrast, fronts dozens of models across OpenAI, Anthropic, Google and xAI behind one endpoint, lets a team bring its own provider keys at zero markup over list price, and adds automatic failover between providers when one is degraded (Vercel, AI Gateway; Vercel docs, Bring Your Own Key). Netlify’s gateway has none of that: no BYOK option, no cross-provider failover, and every model call is metered through the same credit conversion whether or not a team already has its own provider contract. A team that wants Claude Fable 5.1 today has had it on Vercel’s gateway since the model shipped, priced at whatever Anthropic charges (Vercel changelog, Claude Fable 5.1 now available on AI Gateway).
That asymmetry is why Netlify keeps shipping these additions one model at a time. Each one closes a specific “we don’t have that model yet” complaint, but the underlying product — a single-vendor gateway with a fixed markup and no failover — stays unchanged release to release.
The take #
Adding Claude Fable 5.1 and Gemini 3.8 Flash is a reasonable, low-risk move: it keeps AI Gateway from falling further behind on model coverage, and zero-config access saves a team already deployed on Netlify from managing another provider key. But it doesn’t change what AI Gateway is. Vercel built a routing layer that treats providers as interchangeable and prices accordingly; Netlify built a metered pass-through for its own credit plans. A team picking a deploy platform primarily for AI workloads is choosing a gateway architecture, not a model list, and Netlify’s architecture hasn’t moved.