Unify (unify.ai) is a YCombinator W23-backed LLM routing platform founded in 2023 and headquartered in London. Its core product is a unified API gateway that lets developers access dozens of LLM providers with a single API key, and automatically routes each prompt to the model and provider that best satisfies user-defined quality, cost, and latency constraints. The platform's key technical differentiator is its benchmark-driven dynamic router: a neural network trained on Unify's continuously refreshed benchmark suite, which tests models across regions (US, EU, Asia), concurrency levels, and sequence lengths, updating every few minutes. Developers can tune three routing dimensions via simple sliders or programmatic constraints, and the router re-optimises automatically as new models and providers launch. Unify treats all providers neutrally and publishes its benchmarks publicly, covering providers such as OpenAI, Mistral AI, Together AI, Anyscale, Perplexity AI, Replicate, and OctoAI among others. Enterprise users can train custom routers on their own prompt data and deploy fine-tuned smaller models on-premises through Unify's toolkit. By 2025-2026 the product had evolved under the tagline 'Hire AI, Not APIs', expanding toward agentic interfaces where AI colleagues are onboarded like employees (screen walkthroughs, docs, voice notes) and operate in dedicated per-agent computer environments. This layer sits atop the same unified API infrastructure. Pricing is usage-based with a 4% markup on pay-as-you-go, dropping to 0% on annual prepay (Tier 3). Team plans start at $50/month with all features included and no per-seat charges; enterprise plans are custom. Key features: - Single API key access to all major LLM providers (OpenAI, Mistral, Together AI, Anyscale, Replicate, OctoAI, Perplexity, and more) - Benchmark-driven dynamic router (neural network) updated continuously across regions and concurrency levels - Three-dimensional routing control: quality, cost, and latency sliders with programmatic constraint support - Public, unbiased benchmarks updated frequently to reflect real provider performance - Custom router training on user-specific prompt data for enterprise deployments - Fine-tuned model deployment toolkit for on-premises enterprise use - Agentic 'AI colleague' interfaces that operate in dedicated computer environments (2025 expansion) - Team plans from $50/month with all features included; 0% markup on annual prepay
4% markup on pay-as-you-go usage; 0% on Tier 3 annual prepay. Team plans from $50/month (all features, no per-seat charges). Enterprise: custom.