Requesty routes model inference requests to Nebius through its AI gateway. In observations from its production gateway for April 2026, Requesty reported that its Nebius, Minimaxi and DeepInfra routes clustered at 23–26 output tokens per second. Its model catalogue also documents a Nebius-served DeepSeek deployment accessible through Requesty’s OpenAI-compatible API.
Requesty lists Nebius among its inference model subprocessors. Each model provider receives customer data only when requests are routed to a model it operates. Requesty also lists Nebius among the providers available through its Frankfurt routing infrastructure. Its March 2025 guides described Nebius access and optional fallback policies involving Nebius.
For April 2026, Requesty reported that tool-call completions accounted for 0.9% of Nebius output tokens and 3.5% of Nebius requests. Its catalogue reports production traffic for Nebius-served Kimi K3 and GLM-5.2, with median output speeds of 68 and 112 tokens per second respectively. Endpoint-specific instructions for these models and GLM-5.1 describe managed EU deployments with no routing or failover.
Requesty Nebius-hosted DeepSeek model route18 Oct 2026 · Requesty · Web pageRequesty documents a DeepSeek model deployment served by Nebius AI through Requesty’s OpenAI-compatible API.
Requesty subprocessor disclosure
18 Oct 2026 · Requesty · Web page
Requesty lists Nebius AI among its inference model subprocessors in the EU, with no data retention or training on prompts. Model providers process customer data only when requests are routed to models they operate.
Requesty subprocessor disclosure18 Oct 2026 · Requesty · Web pageRequesty lists Nebius AI among its inference model subprocessors in the EU, with no data retention or training on prompts. Model providers process customer data only when requests are routed to models they operate.
Nebius AI glm-5.2 API Pricing & Cost: Context Window & Benchmarks | Requesty
19 Oct 2026 · www.requesty.ai · Web page
Requesty reports measured production traffic for Nebius-served GLM-5.2, with median output speed of 112 tokens per second and median time to first token of 957 milliseconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover.
Nebius AI glm-5.2 API Pricing & Cost: Context Window & Benchmarks | Requesty19 Oct 2026 · www.requesty.ai · Web pageRequesty reports measured production traffic for Nebius-served GLM-5.2, with median output speed of 112 tokens per second and median time to first token of 957 milliseconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover.
Nebius AI kimi-k3 API Pricing & Cost: Context Window & Benchmarks | Requesty
19 Oct 2026 · www.requesty.ai · Web page
Requesty reports measured production traffic for Nebius-served Kimi K3, with median output speed of 68 tokens per second and median time to first token of 1.27 seconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover. Requesty states that pay-as-you-go access adds 5% to upstream rates, while access using customers’ own provider keys adds no percentage.
Nebius AI kimi-k3 API Pricing & Cost: Context Window & Benchmarks | Requesty19 Oct 2026 · www.requesty.ai · Web pageRequesty reports measured production traffic for Nebius-served Kimi K3, with median output speed of 68 tokens per second and median time to first token of 1.27 seconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover. Requesty states that pay-as-you-go access adds 5% to upstream rates, while access using customers’ own provider keys adds no percentage.
Nebius AI zai-org/glm-5.1 API Pricing & Cost: Context Window & Benchmarks | Requesty
19 Oct 2026 · www.requesty.ai · Web page
Requesty documents a Nebius-served GLM-5.1 deployment accessible through its OpenAI-compatible API. Endpoint-specific instructions describe a managed EU deployment with no routing or failover. The deployment metadata lists no data retention and no use for training.
Nebius AI zai-org/glm-5.1 API Pricing & Cost: Context Window & Benchmarks | Requesty19 Oct 2026 · www.requesty.ai · Web pageRequesty documents a Nebius-served GLM-5.1 deployment accessible through its OpenAI-compatible API. Endpoint-specific instructions describe a managed EU deployment with no routing or failover. The deployment metadata lists no data retention and no use for training.
Requesty, operated by UK company Requesty Ltd, provides an AI gateway for developers and enterprise teams building applications with multiple model providers. Its OpenAI-compatible API combines model access with routing, automatic failover, caching and usage analytics. Teams can track costs, set spending limits and manage API keys through a shared platform. Enterprise features include single sign-on, role-based access, approved-model policies and EU routing through Frankfurt, helping organizations manage access and governance across their AI applications.
Requesty partners with Melious for European model inference
Requesty integrated Melious into its gateway, allowing customers to run open-weight models on hardware in European datacenters through their existing Requesty API integration.
DoiT announced a Cloud Intelligence integration that imports Requesty usage and spending by model, hosting provider, API key and routing policy. It supports reporting, budgets, anomaly detection and forecasts alongside other cloud spending.
Requesty provides an AI gateway with model routing, failover, analytics and governance through an OpenAI-compatible API. · Infrastructure
How they use Nebius Token Factory
Requesty routes model inference requests to Nebius through its AI gateway. In observations from its production gateway for April 2026, Requesty reported that its Nebius, Minimaxi and DeepInfra routes clustered at 23–26 output tokens per second. Its model catalogue also documents a Nebius-served DeepSeek deployment accessible through Requesty’s OpenAI-compatible API.
Requesty lists Nebius among its inference model subprocessors. Each model provider receives customer data only when requests are routed to a model it operates. Requesty also lists Nebius among the providers available through its Frankfurt routing infrastructure. Its March 2025 guides described Nebius access and optional fallback policies involving Nebius.
For April 2026, Requesty reported that tool-call completions accounted for 0.9% of Nebius output tokens and 3.5% of Nebius requests. Its catalogue reports production traffic for Nebius-served Kimi K3 and GLM-5.2, with median output speeds of 68 and 112 tokens per second respectively. Endpoint-specific instructions for these models and GLM-5.1 describe managed EU deployments with no routing or failover.
Requesty Nebius-hosted DeepSeek model route18 Oct 2026 · Requesty · Web pageRequesty documents a DeepSeek model deployment served by Nebius AI through Requesty’s OpenAI-compatible API.
Requesty subprocessor disclosure
18 Oct 2026 · Requesty · Web page
Requesty lists Nebius AI among its inference model subprocessors in the EU, with no data retention or training on prompts. Model providers process customer data only when requests are routed to models they operate.
Requesty subprocessor disclosure18 Oct 2026 · Requesty · Web pageRequesty lists Nebius AI among its inference model subprocessors in the EU, with no data retention or training on prompts. Model providers process customer data only when requests are routed to models they operate.
Nebius AI glm-5.2 API Pricing & Cost: Context Window & Benchmarks | Requesty
19 Oct 2026 · www.requesty.ai · Web page
Requesty reports measured production traffic for Nebius-served GLM-5.2, with median output speed of 112 tokens per second and median time to first token of 957 milliseconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover.
Nebius AI glm-5.2 API Pricing & Cost: Context Window & Benchmarks | Requesty19 Oct 2026 · www.requesty.ai · Web pageRequesty reports measured production traffic for Nebius-served GLM-5.2, with median output speed of 112 tokens per second and median time to first token of 957 milliseconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover.
Nebius AI kimi-k3 API Pricing & Cost: Context Window & Benchmarks | Requesty
19 Oct 2026 · www.requesty.ai · Web page
Requesty reports measured production traffic for Nebius-served Kimi K3, with median output speed of 68 tokens per second and median time to first token of 1.27 seconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover. Requesty states that pay-as-you-go access adds 5% to upstream rates, while access using customers’ own provider keys adds no percentage.
Nebius AI kimi-k3 API Pricing & Cost: Context Window & Benchmarks | Requesty19 Oct 2026 · www.requesty.ai · Web pageRequesty reports measured production traffic for Nebius-served Kimi K3, with median output speed of 68 tokens per second and median time to first token of 1.27 seconds. Its endpoint-specific instructions describe a managed EU deployment with no routing or failover. Requesty states that pay-as-you-go access adds 5% to upstream rates, while access using customers’ own provider keys adds no percentage.
Nebius AI zai-org/glm-5.1 API Pricing & Cost: Context Window & Benchmarks | Requesty
19 Oct 2026 · www.requesty.ai · Web page
Requesty documents a Nebius-served GLM-5.1 deployment accessible through its OpenAI-compatible API. Endpoint-specific instructions describe a managed EU deployment with no routing or failover. The deployment metadata lists no data retention and no use for training.
Nebius AI zai-org/glm-5.1 API Pricing & Cost: Context Window & Benchmarks | Requesty19 Oct 2026 · www.requesty.ai · Web pageRequesty documents a Nebius-served GLM-5.1 deployment accessible through its OpenAI-compatible API. Endpoint-specific instructions describe a managed EU deployment with no routing or failover. The deployment metadata lists no data retention and no use for training.
Requesty, operated by UK company Requesty Ltd, provides an AI gateway for developers and enterprise teams building applications with multiple model providers. Its OpenAI-compatible API combines model access with routing, automatic failover, caching and usage analytics. Teams can track costs, set spending limits and manage API keys through a shared platform. Enterprise features include single sign-on, role-based access, approved-model policies and EU routing through Frankfurt, helping organizations manage access and governance across their AI applications.
Requesty partners with Melious for European model inference
Requesty integrated Melious into its gateway, allowing customers to run open-weight models on hardware in European datacenters through their existing Requesty API integration.
DoiT announced a Cloud Intelligence integration that imports Requesty usage and spending by model, hosting provider, API key and routing policy. It supports reporting, budgets, anomaly detection and forecasts alongside other cloud spending.
