All promotions
GuidesMCP

Nscale

Inference providerUnited Kingdomprices checked Oct 23, 12:30 UTC

Pricing page ↗

Nscale serves 9 models at a verified price. It is the cheapest host on 5 of the 9 it shares with others; its median price matches the cheapest.

Models served97 with open weights
Cheapest input, per M$0.05gpt-oss-20b
Sign-upno free credits stated
Newest first
ModelPrecisionContextToolsInputOutputCache
gpt-oss-120bnot read: host terms128K—$0.10$0.40—
gpt-oss-20bnot read: host terms128K—$0.05$0.20—
Qwen3 235B A22B Instruct 2507not read: host terms256K—$0.20$0.60—
Qwen3 32Bcheapestnot read: host terms40K—$0.08$0.25—
Qwen3 8Bcheapest—128K—$0.07$0.18—
Qwen3 14Bcheapestnot read: host terms40K—$0.07$0.20—
Llama 4 Scoutcheapestnot read: host terms320K—$0.09$0.29—
Qwen2.5 Coder 32B Instructcheapest—32K—$0.06$0.20—
Llama 3.1 8Bnot read: host terms128K—$0.06$0.06—

9 of 9 models · prices per million tokens · precision and context: read at Nscale's own catalogue or docs where its terms allow, otherwise as declared to OpenRouter (hover a cell for its source) · tools as Nscale declares them to OpenRouter

ConditionWhat it saysSource
Sign-upNot checked yetdocs.nscale.com
Free creditsNone stated—
Headquarters regionUnited Kingdom—
Price listWhere its per-token prices are publishednscale.com
API documentationIts own docsdocs.nscale.com
Terms of serviceIts termsnscale.com

Links checked Oct 23.