flo2.com

What Is It

flo2 is an LLM Gateway, Router, and Proxy built for developers. It does not resell tokens; instead, users bring their own provider keys from OpenAI, Anthropic, Groq, Cerebras, DeepInfra, and others, then route requests through a single flo2 Key. Its positioning is close to an OpenRouter alternative, but with an emphasis on zero markup, BYOK, and private by default.

Core Capabilities

Functionally, flo2 is not trying to build yet another model, but to manage routing across multiple model providers. It supports intelligent routing, allowing requests to be directed to a default model, a restricted set of models, or an entire key pool. Fallback chains can automatically switch to backup models after the primary model fails and retries are exhausted. AI racing sends requests to multiple models in parallel and uses the fastest response. A/B testing can run shadow tests on new models and evaluate outputs with a judge model. On the cost side, flo2 records tokens, throughput, and estimated compute cost for each call, and provides dashboards broken down by model. Its API is compatible with OpenAI Chat Completions, Responses, legacy Completions, and Anthropic Messages, and it supports streaming output; migration mainly involves replacing the base URL.

Pricing and Privacy

The current page clearly states that it is free during the Beta period and that there is no markup on tokens; users still pay model providers directly. The long-term business model has not been disclosed, which makes budget planning uncertain. In terms of privacy, flo2 does not log prompts or responses by default, retaining only metadata such as tokens, latency, and cost. Content is stored only when users enable A/B testing or Prompt Insights, and is deleted after a short retention window.

Pros and Cons

The advantages are low integration cost, compatibility with mainstream APIs, and practical support for multi-model failover and cost optimization, making it especially suitable for high-volume LLM workloads. The drawbacks are that it is still in Beta, with no clear information on SLA, rate limits, enterprise support, or long-term pricing. The page also does not indicate whether open-source, self-hosted, or private deployment options are available. For teams with strict compliance requirements, a hosted gateway still needs additional security assessment.

Who It’s For and Access from China

flo2 is suitable for teams building AI applications, research tools, data APIs, content-processing pipelines, and other systems that already generate substantial LLM traffic and want to reduce costs while avoiding single-vendor failures. The page does not specify access from mainland China, payment methods, or local compliance status, so china_access can only be considered unknown. If access or payment is restricted, alternatives include integrating directly with model providers, using OpenRouter, or building a self-hosted LLM Gateway.