TrustedRouter uses Nebius Token Factory as a downstream inference provider. Its provider catalog has listed active routing and 21 public models. Its October 2026 EU catalog lists 26 configured Nebius Token Factory models. Its subprocessor disclosure says Nebius processes prompt and output content for requests routed to it. TrustedRouter continuously samples routed Nebius models to measure time to first token, effective throughput and success rate without storing prompt or output content. Its MiniCPM-V-4_5 model page lists a Nebius Token Factory route and reports 1,143 ms median time to first token and 100% recent uptime from routed probes. TrustedRouter offers developers and businesses one OpenAI-compatible API with controls for price, reliability, region and privacy. Its gpt-oss-120b page also lists a Nebius Token Factory provider route with token pricing and a zero data retention designation. Its Hermes 4 405B page lists Nebius Token Factory as its sole provider route with a zero data retention designation.
TrustedRouter states that zero retention is an opt-in control enabled on its deployed Nebius account. Its catalog says Nebius retains inputs and outputs by default for speculative decoding. Nebius does not train on customer data. The content-retention claim excludes account and billing metadata. TrustedRouter does not mark the Nebius provider boundary as end-to-end encrypted. Its disclosure separately notes that opt-in provider-native Batch may temporarily retain plaintext inputs for up to 26 hours and results for up to 6 hours after availability. Requests can require zero retention by setting provider.min_privacy to zdr. Routing fails closed if no eligible route remains. TrustedRouter documents no verified provider-specific location pin for Nebius.
Its June 2026 benchmark reports 13,781 Nebius observations, 97.37% measured availability, 1,540 ms median time to first token and 8,021 ms p95 time to first token. The report includes organic traffic observations and synthetic route probes. Its DeepSeek V4 Pro page reports routed probes through Nebius Token Factory with 2,655 ms median time to first token and 100% recent route uptime. Its Kimi K3 page reports 991 ms median time to first token through Nebius Token Factory, the fastest measured provider route on that page. Its Qwen3 30B A3B Instruct 2507 page reports 1,331 ms median time to first token and 100% recent route uptime from routed probes through Nebius Token Factory.
TrustedRouter lists active routing for 24 public Nebius Token Factory models and measures routed model performance without storing prompt or output content. Its provider page documents no verified inference location, country-specific residency commitment or provider-specific location pin through TrustedRouter.
Nebius Token Factory Models, Pricing & Privacy7 Oct 2026 · trustedrouter.com · Web pageTrustedRouter lists active routing for 24 public Nebius Token Factory models and measures routed model performance without storing prompt or output content. Its provider page documents no verified inference location, country-specific residency commitment or provider-specific location pin through TrustedRouter.
Qwen3 30B A3B Instruct 2507 API, Pricing and Providers
18 Oct 2026 · trustedrouter.com · Web page
TrustedRouter lists Nebius Token Factory as its sole route for Qwen3 30B A3B Instruct 2507. It reports 1,229 ms median time to first token and 100% recent route uptime from routed probes. Requests can require zero retention using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
Qwen3 30B A3B Instruct 2507 API, Pricing and Providers18 Oct 2026 · trustedrouter.com · Web pageTrustedRouter lists Nebius Token Factory as its sole route for Qwen3 30B A3B Instruct 2507. It reports 1,229 ms median time to first token and 100% recent route uptime from routed probes. Requests can require zero retention using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
Subprocessors | TrustedRouter
1 Oct 2026 · TrustedRouter · Web page
TrustedRouter identifies Nebius Token Factory as a downstream inference provider that processes prompt and output content for routed requests. Its deployed Nebius account has opt-in zero retention enabled. The disclosure says opt-in provider-native Batch may retain plaintext inputs for up to 26 hours and results for up to 6 hours after availability.
Subprocessors | TrustedRouter1 Oct 2026 · TrustedRouter · Web pageTrustedRouter identifies Nebius Token Factory as a downstream inference provider that processes prompt and output content for routed requests. Its deployed Nebius account has opt-in zero retention enabled. The disclosure says opt-in provider-native Batch may retain plaintext inputs for up to 26 hours and results for up to 6 hours after availability.
openbmb/MiniCPM-V-4_5 API, Pricing and Providers
14 Oct 2026 · trustedrouter.com · Web page
TrustedRouter reports 1,143 ms median time to first token and 100% recent uptime from routed MiniCPM-V-4_5 probes through Nebius Token Factory.
openbmb/MiniCPM-V-4_5 API, Pricing and Providers14 Oct 2026 · trustedrouter.com · Web pageTrustedRouter reports 1,143 ms median time to first token and 100% recent uptime from routed MiniCPM-V-4_5 probes through Nebius Token Factory.
GLM-5.2 API, Pricing and Providers | TrustedRouter
15 Oct 2026 · trustedrouter.com · Web page
TrustedRouter lists Nebius Token Factory as its sole GLM-5.2 provider route and reports 100% recent route uptime from routed probes. Requests can require recorded zero-retention routes using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
GLM-5.2 API, Pricing and Providers | TrustedRouter15 Oct 2026 · trustedrouter.com · Web pageTrustedRouter lists Nebius Token Factory as its sole GLM-5.2 provider route and reports 100% recent route uptime from routed probes. Requests can require recorded zero-retention routes using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
TrustedRouter, operated by Lore Hex Corp, provides an OpenAI-compatible API that connects developers and businesses to AI models across multiple providers. Its main product is an open-source model gateway with controls for pricing, provider fallback, regional routing and privacy requirements, including zero-retention and confidential-compute routes. Teams use it for coding agents, document processing, research, evaluations and applications that need multiple model providers. The service also offers Python and JavaScript SDKs, model and provider catalogs, and hardware-attestation tools for checking the gateway’s running software.
TrustedRouter raises $6 million seed round led by Sierra Ventures
TrustedRouter told Semafor it closed a $6 million seed round that week, led by Sierra Ventures. The company provides AI model routing with confidential computing options.
TrustedRouter introduced an API that returns typed decisions and probabilities from a supplied state and named questions. It supports TypeSafe AI’s Jev and eight additional model configurations through the existing gateway, API key and billing system.
TrustedRouter raises $1.25 million in seed funding
TrustedRouter announced a $1.25 million seed round to develop and grow its open-source AI router. Investors include Sam Lessin of Slow Ventures, Bill Tai, Linda Avey and George Xing.
TrustedRouter provides an OpenAI-compatible API for AI models served across multiple providers. · Infrastructure
How they use Nebius Token Factory
TrustedRouter uses Nebius Token Factory as a downstream inference provider. Its provider catalog has listed active routing and 21 public models. Its October 2026 EU catalog lists 26 configured Nebius Token Factory models. Its subprocessor disclosure says Nebius processes prompt and output content for requests routed to it. TrustedRouter continuously samples routed Nebius models to measure time to first token, effective throughput and success rate without storing prompt or output content. Its MiniCPM-V-4_5 model page lists a Nebius Token Factory route and reports 1,143 ms median time to first token and 100% recent uptime from routed probes. TrustedRouter offers developers and businesses one OpenAI-compatible API with controls for price, reliability, region and privacy. Its gpt-oss-120b page also lists a Nebius Token Factory provider route with token pricing and a zero data retention designation. Its Hermes 4 405B page lists Nebius Token Factory as its sole provider route with a zero data retention designation.
TrustedRouter states that zero retention is an opt-in control enabled on its deployed Nebius account. Its catalog says Nebius retains inputs and outputs by default for speculative decoding. Nebius does not train on customer data. The content-retention claim excludes account and billing metadata. TrustedRouter does not mark the Nebius provider boundary as end-to-end encrypted. Its disclosure separately notes that opt-in provider-native Batch may temporarily retain plaintext inputs for up to 26 hours and results for up to 6 hours after availability. Requests can require zero retention by setting provider.min_privacy to zdr. Routing fails closed if no eligible route remains. TrustedRouter documents no verified provider-specific location pin for Nebius.
Its June 2026 benchmark reports 13,781 Nebius observations, 97.37% measured availability, 1,540 ms median time to first token and 8,021 ms p95 time to first token. The report includes organic traffic observations and synthetic route probes. Its DeepSeek V4 Pro page reports routed probes through Nebius Token Factory with 2,655 ms median time to first token and 100% recent route uptime. Its Kimi K3 page reports 991 ms median time to first token through Nebius Token Factory, the fastest measured provider route on that page. Its Qwen3 30B A3B Instruct 2507 page reports 1,331 ms median time to first token and 100% recent route uptime from routed probes through Nebius Token Factory.
TrustedRouter lists active routing for 24 public Nebius Token Factory models and measures routed model performance without storing prompt or output content. Its provider page documents no verified inference location, country-specific residency commitment or provider-specific location pin through TrustedRouter.
Nebius Token Factory Models, Pricing & Privacy7 Oct 2026 · trustedrouter.com · Web pageTrustedRouter lists active routing for 24 public Nebius Token Factory models and measures routed model performance without storing prompt or output content. Its provider page documents no verified inference location, country-specific residency commitment or provider-specific location pin through TrustedRouter.
Qwen3 30B A3B Instruct 2507 API, Pricing and Providers
18 Oct 2026 · trustedrouter.com · Web page
TrustedRouter lists Nebius Token Factory as its sole route for Qwen3 30B A3B Instruct 2507. It reports 1,229 ms median time to first token and 100% recent route uptime from routed probes. Requests can require zero retention using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
Qwen3 30B A3B Instruct 2507 API, Pricing and Providers18 Oct 2026 · trustedrouter.com · Web pageTrustedRouter lists Nebius Token Factory as its sole route for Qwen3 30B A3B Instruct 2507. It reports 1,229 ms median time to first token and 100% recent route uptime from routed probes. Requests can require zero retention using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
Subprocessors | TrustedRouter
1 Oct 2026 · TrustedRouter · Web page
TrustedRouter identifies Nebius Token Factory as a downstream inference provider that processes prompt and output content for routed requests. Its deployed Nebius account has opt-in zero retention enabled. The disclosure says opt-in provider-native Batch may retain plaintext inputs for up to 26 hours and results for up to 6 hours after availability.
Subprocessors | TrustedRouter1 Oct 2026 · TrustedRouter · Web pageTrustedRouter identifies Nebius Token Factory as a downstream inference provider that processes prompt and output content for routed requests. Its deployed Nebius account has opt-in zero retention enabled. The disclosure says opt-in provider-native Batch may retain plaintext inputs for up to 26 hours and results for up to 6 hours after availability.
openbmb/MiniCPM-V-4_5 API, Pricing and Providers
14 Oct 2026 · trustedrouter.com · Web page
TrustedRouter reports 1,143 ms median time to first token and 100% recent uptime from routed MiniCPM-V-4_5 probes through Nebius Token Factory.
openbmb/MiniCPM-V-4_5 API, Pricing and Providers14 Oct 2026 · trustedrouter.com · Web pageTrustedRouter reports 1,143 ms median time to first token and 100% recent uptime from routed MiniCPM-V-4_5 probes through Nebius Token Factory.
GLM-5.2 API, Pricing and Providers | TrustedRouter
15 Oct 2026 · trustedrouter.com · Web page
TrustedRouter lists Nebius Token Factory as its sole GLM-5.2 provider route and reports 100% recent route uptime from routed probes. Requests can require recorded zero-retention routes using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
GLM-5.2 API, Pricing and Providers | TrustedRouter15 Oct 2026 · trustedrouter.com · Web pageTrustedRouter lists Nebius Token Factory as its sole GLM-5.2 provider route and reports 100% recent route uptime from routed probes. Requests can require recorded zero-retention routes using provider.min_privacy set to zdr, with routing failing closed if no eligible route remains.
TrustedRouter, operated by Lore Hex Corp, provides an OpenAI-compatible API that connects developers and businesses to AI models across multiple providers. Its main product is an open-source model gateway with controls for pricing, provider fallback, regional routing and privacy requirements, including zero-retention and confidential-compute routes. Teams use it for coding agents, document processing, research, evaluations and applications that need multiple model providers. The service also offers Python and JavaScript SDKs, model and provider catalogs, and hardware-attestation tools for checking the gateway’s running software.
TrustedRouter raises $6 million seed round led by Sierra Ventures
TrustedRouter told Semafor it closed a $6 million seed round that week, led by Sierra Ventures. The company provides AI model routing with confidential computing options.
TrustedRouter introduced an API that returns typed decisions and probabilities from a supplied state and named questions. It supports TypeSafe AI’s Jev and eight additional model configurations through the existing gateway, API key and billing system.
TrustedRouter raises $1.25 million in seed funding
TrustedRouter announced a $1.25 million seed round to develop and grow its open-source AI router. Investors include Sam Lessin of Slow Ventures, Bill Tai, Linda Avey and George Xing.
