Cloudflare Auto Router hits public beta, claims up to 30% AI cost savings

Cloudflare has released its Auto Router in public beta through AI Gateway. Set your model to cloudflare/auto and the router automatically sends each request to a model it judges capable enough for the task, so users don’t pick a model themselves.

The company says its early internal results, run through its OpenCode harness, show cost savings of up to 30% compared with using only frontier models such as OpenAI Sol and Anthropic Claude Opus.

What the internal benchmark found

Cloudflare says it evaluated cloudflare/auto against OpenAI’s GPT-6 Sol and Anthropic’s Claude Opus 5.5 on its internal general knowledge work benchmark, which uses simulated workspace tools and covers day-to-day workflows across email, calendars, Slack, files, travel and finance. Each task requires the model to use those tools to produce a verifiable answer or complete an action.

According to the company’s numbers, cloudflare/auto succeeded on 252 of 291 trials, an 86.6% success rate, at a total cost of $2.10 — $0.0084 per success. Claude Opus 5.5 succeeded on 281 of 291 trials (96.6%) for $5.91, or $0.0210 per success, while GPT-6 Sol succeeded on 245 of 291 (84.2%) for $2.64, or $0.0108 per success. Cloudflare says the benchmark covered 97 tasks with three samples per model per task, and that the parenthetical confidence intervals were estimated from 10,000 task-level bootstrap resamples.

Cloudflare also says it uses Auto Router internally within its OpenCode deployment and within Cloudflare OS, its custom agent harness, and that internal usage produced results comparable with frontier models for coding tasks.

Why the gateway makes the call

The company’s argument is that in many harnesses, including OpenCode, Claude Code and Codex, users still select models manually, and often reach for models that are overkill — summarizing an email or a chat thread doesn’t need Opus-level intelligence, Cloudflare notes, though you wouldn’t want to block that model entirely from a security engineering team.

Because every request from every user, agent and tool already flows through AI Gateway, Cloudflare says the gateway is positioned to do more than observe and enforce. Budgets, spend limits and identity-aware analytics give organizations visibility and guardrails, but still rely on individuals to make cost-conscious choices request by request, according to the company. The next step, in Cloudflare’s framing, is for the gateway to make those decisions on a user’s behalf.

Cloudflare says Auto Router performs best across a wide range of knowledge-work tasks, like those in a large organization with both technical and non-technical teams.

Auto Router is available now in public beta through AI Gateway. To try it, set your model to cloudflare/auto.

Related reading

Sources

• Cut your AI spend with AI Gateway's Auto Router — Cloudflare Blog