Nscale
Try Nscale on us. Starter credit covers your first requests, then one key reaches all 20 Nscale models, with the privacy posture shown on every route.
Nscalenscale
These privacy labels describe Nscale, the upstream model provider. Confidential requires all three: verified provider compute, end-to-end encryption and explicit zero data retention. Account and billing metadata are outside the content-retention claim.
| Provider | Nscale |
|---|---|
| API operator country | GB Legal home of the API operator, not an inference-location guarantee. |
| Headquarters | Not verified |
| Routing status | Active |
| Provider website | https://www.nscale.com/ |
| Models | 20 public models |
| Credits routes | 20 |
| Zero data retention | no |
| Verified confidential inference | Not verified |
| Policy note | Nscale publishes a general nonlogging statement for serverless inference, but TrustedRouter has not verified a contractual zero-retention commitment for this account. These routes remain Standard and make no confidential-compute or E2EE claim. Policy source |
Inference locations
Provider API declarations are also available on each model's endpoint API. Direct chat and Responses replies separate advertised availability from serving location in usage.inference_location; the actual serving region remains unknown when not reported.
usage.provider_usage.gateway_region identifies the TrustedRouter gateway, not the provider's inference location. A value such as europe-west4 does not mean the model ran in Europe. usage.inference_location.serving_region remains null when the provider has not reported it. Older gateways may return the gateway-only field as provider_usage.region.
Where the upstream GPUs execute the model, not the provider's headquarters, API ingress, or data storage. ZDR and confidential inference do not by themselves restrict geography.
- Declared countries or regions
- Not verified
No inference-location declaration has been verified for this provider. - Can locations change?
- Not verified. Do not assume that the serving location is fixed.
- Provider region pinning
- Not verified.
- Pinning through TrustedRouter
- No verified provider-specific location pin documented here.
- Infrastructure declaration
- Inference location not verified; no country-specific residency commitment.
- Evidence
- Not yet reviewed.
Provider selection alone is not a region pin. TrustedRouter's US provider filter selects company jurisdiction, not GPU location. A regional gateway hostname does not pin a downstream GPU. For a binding residency requirement, confirm the model, region, failover scope and enforced controls before sending sensitive data.
Measured performance
29 samplesContinuously sampled across Nscale's routed models: p50 TTFT, effective throughput, and success rate. Effective throughput uses provider-reported output tokens over complete request time. Unsupported route and probe-configuration rows are separated from provider downtime. No prompt or output content stored.
| p50 TTFT | 1311 ms |
|---|---|
| Effective throughput | 127 tok/s n=1 |
| Uptime | 100.00% |
| Model | p50 TTFT | Effective throughput | Uptime | Config excluded | Availability samples |
|---|---|---|---|---|---|
| nvidia/nemotron-3-nano-30b-a3b-bf16 | 1115 ms | — | 100.00% | — | 2 |
| deepseek/deepseek-r1-distill-qwen-14b | 1181 ms | — | 100.00% | — | 2 |
| openai/gpt-oss-120b | 1209 ms | — | 100.00% | — | 2 |
| qwen/qwen3-32b | 1248 ms | — | 100.00% | — | 4 |
| qwen/qwen2.5-coder-32b-instruct | 1305 ms | — | 100.00% | — | 1 |
| qwen/qwen3-4b-instruct-2507 | 1311 ms | — | 100.00% | — | 5 |
| mistralai/devstral-small-2505 | 1377 ms | — | 100.00% | — | 1 |
| qwen/qwen2.5-coder-7b-instruct | 1378 ms | — | 100.00% | — | 2 |
| qwen/qwen3-14b | 1454 ms | — | 100.00% | — | 1 |
| qwen/qwen3-235b-a22b-instruct-2507 | 1484 ms | — | 100.00% | — | 1 |
| moonshotai/kimi-k2.5 | 1506 ms | — | 100.00% | — | 1 |
| z-ai/glm-5.3 | 1701 ms | 127 tok/s n=1 | 100.00% | — | 2 |
| meta-llama/llama-3.1-8b-instruct | 1777 ms | — | 100.00% | — | 3 |
| qwen/qwen3.8-flash-next-fp8 | 1890 ms | — | 100.00% | — | 2 |
Nscale performance history · Full provider & model leaderboard.
Models served by Nscale.
Each row links to pricing, provider, benchmark, and API pages for the model.
| Model | AI IQ | Context | Input | Cached input | Output |
|---|---|---|---|---|---|
black-forest-labs/flux.1-schnellFLUX.1 Schnell |
— | 0 | selected route | Not published | selected route |
deepseek/deepseek-r1-distill-qwen-14bDeepSeek-R1-Distill-Qwen-14B |
— | 131,072 | $0.211/1M | Not published | $0.211/1M |
meta-llama/llama-3.1-8b-instructMeta: Llama 3.1 8B Instruct |
— | 131,072 | $0.0633/1M | Not published | $0.0633/1M |
mistralai/devstral-small-2505Devstral-Small-2505 |
— | 128,000 | $0.1055/1M | Not published | $0.3165/1M |
moonshotai/kimi-k2.5MoonshotAI: Kimi K2.5 |
IQ 110#82 | 262,144 | $0.47475/1M | Not published | $2.321/1M |
nvidia/nemotron-3-nano-30b-a3b-bf16NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 |
— | 256,000 | $0.05275/1M | Not published | $0.211/1M |
openai/gpt-oss-120bOpenAI: gpt-oss-120b |
IQ 98#136 | 131,072 | $0.1055/1M | Not published | $0.422/1M |
openai/gpt-oss-20bOpenAI: gpt-oss-20b |
IQ 93#155 | 131,072 | $0.05275/1M | Not published | $0.211/1M |
qwen/qwen2.5-coder-32b-instruct@cf/qwen/qwen2.5-coder-32b-instruct |
— | 32,768 | $0.0633/1M | Not published | $0.211/1M |
qwen/qwen2.5-coder-3b-instructQwen2.5-Coder-3B-Instruct |
— | 32,768 | $0.01055/1M | Not published | $0.03165/1M |
qwen/qwen2.5-coder-7b-instructQwen2.5-Coder-7B-Instruct |
— | 131,072 | $0.01055/1M | Not published | $0.03165/1M |
qwen/qwen3-14bQwen: Qwen3 14B |
— | 40,960 | $0.07385/1M | Not published | $0.211/1M |
qwen/qwen3-235b-a22bQwen3-235B-A22B |
— | 32,768 | $0.211/1M | Not published | $0.633/1M |
qwen/qwen3-235b-a22b-instruct-2507Qwen/Qwen3-235B-A22B-Instruct-2507 |
— | 262,144 | $0.211/1M | Not published | $0.633/1M |
qwen/qwen3-32bQwen: Qwen3 32B |
— | 131,072 | $0.0844/1M | Not published | $0.26375/1M |
qwen/qwen3-4b-instruct-2507Qwen3-4B-Instruct-2507 |
— | 262,144 | $0.01055/1M | Not published | $0.03165/1M |
qwen/qwen3-embedding-8bQwen3 Embedding 8B |
— | 32,768 | $0.0422/1M | Not published | selected route |
qwen/qwen3.8-27bQwen: Qwen3.8 27B |
IQ 115#57 | 65,536 | $0.211/1M | Not published | $2.6375/1M |
qwen/qwen3.8-flash-next-fp8Qwen3.8-Flash-Next-FP8 |
— | 1,048,576 | $0.15825/1M | Not published | $0.49585/1M |
z-ai/glm-5.3Z.ai: GLM 5.3 |
IQ 124#30 | 1,048,576 | $1.477/1M | Not published | $4.642/1M |
Questions
Does Nscale have zero data retention?
TrustedRouter does not currently mark Nscale as provider-level zero data retention. Use trustedrouter/zdr or provider.min_privacy=zdr to select a different eligible route, and review the linked policy source for changes.
Is Nscale end-to-end encrypted?
TrustedRouter does not currently mark Nscale as end-to-end encrypted at the provider boundary. The TrustedRouter gateway is still attested, but the selected provider normally receives the request in order to run the model. Use trustedrouter/e2e for the stronger route requirement.
Which Nscale models are available through TrustedRouter?
This page currently lists 20 public Nscale models, with live pricing, route count, context length, measured performance when available, and links to each model's provider and benchmark pages.