Prem

Declared: Attested E2EE

OpenAI-compatible end-to-end encrypted confidential inference on dedicated GPUs

Prem (prem.io) offers OpenAI-compatible confidential inference where the Confidential Proxy encrypts requests end-to-end to the enclave (XWing key encapsulation + XChaCha20-Poly1305). Attestation combines AMD SEV-SNP or Intel TDX CPU evidence with NVIDIA GPU signed EAT tokens, verified client-side by the open-source Reticle verifier. The service is in beta; the confidential catalog is served per API key from the gateway at gateway.prem.io and is distinct from the non-TEE Zero Data Retention (ZDR) catalog.

Models5

Input from$0.30/M input tokens

Organization rolesStorefront · Workload operator · TEE platform

E2EE (declared)XWING-XCHACHA20-POLY1305

Advertised ChannelAttested E2EE

Post-Quantum KEXHybrid TLS observedX25519MLKEM768 · measured 2026-10-17 · PQC run #31

API Features

OpenAI Compatible✓

Streaming✓

Function Calling✓

Batch Inference—

Vision✓

Embeddings—

Attestation Status

View full details →

Latest attempt: Checked · observed Oct 17, 2026, 5:35 PM

Latest attempt complete; discovery complete; 5/5 targets completed

Model coverage: 5/5 current catalog models have an explicit check.

Availability issueThe verifier could not fetch fresh evidence for every check in this run.

0 verified2 incomplete0 failed3 unavailable

Hardware3 warningTCB policy3 warningFreshness3 warningChannel binding5 warningWorkload identity5 warningSource provenance5 warningRequest route5 warningAvailability3 warning

Endpoint unavailable 3Channel binding not proven 2Workload identity not proven 2Source provenance not proven 2

◐DeepSeek V4.1 Flash (TEE)Intel TDXTEE-only Gatewayprem · declaredDetails ⚠General Nova 3 (TEE)prem · declaredDetails ◐Nemotron 3 Ultra 550B A55B (TEE)Intel TDXTEE-only Gatewayprem · declaredDetails ⚠Qwen36 27B (TEE)prem · declaredDetails ⚠Qwen38 27B (TEE)prem · declaredDetails

Get notified when this provider's attestation verdict changes:Alerts RSS

Models & Pricing

Rates are exactly what the provider source published, kept in the recorded currency; USD eligibility is evaluated from fresh observations. Declared conditions and undisclosed terms are listed, not verified.

Model / source ID	Family	Context	Max Out	Offerings	Attestation	Observed route
DeepSeek V4.1 Flashdeepseek-v4.1-flash	DeepSeek	—	—	per tokens`deepseek-v4.1-flash#ch` In $0.35 /M · out $1.40 /M fresh live price source · observed 2026-10-17 17:35:08Z · price source Per-1M-token USD rates are read live from GET /rvenc/models (price_config.confidential, per-1K fields x 1000). Confidential cached-prompt pricing is a promotional $0.00/1M until 2026-10-20 per Prem docs; records without a cached rate keep none — a post-promo rate is not assumed. Reasoning-token rates are surfaced in offer notes.	incomplete	premdeclared · route not bound
Qwen 3.8 27Bqwen38-27b	Qwen	—	—	per tokens`qwen38-27b#ch` In $0.30 /M · out $2.40 /M Cached read $0 /M · write — /M fresh live price source · observed 2026-10-17 17:35:08Z · price source Per-1M-token USD rates are read live from GET /rvenc/models (price_config.confidential, per-1K fields x 1000). Confidential cached-prompt pricing is a promotional $0.00/1M until 2026-10-20 per Prem docs; records without a cached rate keep none — a post-promo rate is not assumed. Reasoning-token rates are surfaced in offer notes.; reasoning tokens $1.2/1M	unreachable	premdeclared · route not bound
Nemotron 3 Ultranvidia/nemotron-3-ultra-550b-a55b	NVIDIA Nemotron	—	—	per tokens`nvidia/nemotron-3-ultra-550b-a55b#ch` In $0.625 /M · out $3.125 /M Cached read $0 /M · write — /M fresh live price source · observed 2026-10-17 17:35:08Z · price source Per-1M-token USD rates are read live from GET /rvenc/models (price_config.confidential, per-1K fields x 1000). Confidential cached-prompt pricing is a promotional $0.00/1M until 2026-10-20 per Prem docs; records without a cached rate keep none — a post-promo rate is not assumed. Reasoning-token rates are surfaced in offer notes.	incomplete	premdeclared · route not bound
Qwen 3.6qwen36-27b	Qwen	—	—	per tokens`qwen36-27b#ch` In $0.30 /M · out $2.40 /M Cached read $0 /M · write — /M fresh live price source · observed 2026-10-17 17:35:08Z · price source Per-1M-token USD rates are read live from GET /rvenc/models (price_config.confidential, per-1K fields x 1000). Confidential cached-prompt pricing is a promotional $0.00/1M until 2026-10-20 per Prem docs; records without a cached rate keep none — a post-promo rate is not assumed. Reasoning-token rates are surfaced in offer notes.; reasoning tokens $1.2/1M	unreachable	premdeclared · route not bound
DeepGram General Nova 3deepgram/general-nova-3	DeepGram	—	—	per audio-minute`deepgram/general-nova-3#ch` $0.05 per audio minute fresh live price source · observed 2026-10-17 17:35:08Z · price source Per-1M-token USD rates are read live from GET /rvenc/models (price_config.confidential, per-1K fields x 1000). Confidential cached-prompt pricing is a promotional $0.00/1M until 2026-10-20 per Prem docs; records without a cached rate keep none — a post-promo rate is not assumed. Reasoning-token rates are surfaced in offer notes.	unreachable	premdeclared · route not bound