NEWMEN
← back to all models

DeepSeek: R1 0528

deepseek/deepseek-r1-0528

DeepSeekopen-weight164K context5 providersIntelligence 81.9

Cheapest provider

$0.50 / 1M

DeepInfra

Fastest provider (p95)

76 tok/s

DeepInfra

Intelligence (composite)

81.9

MMLU-Pro · HumanEval · math · GPQA

Per-provider performance

Latency / throughput / uptime / price measured across providers over the last 30 minutes of live traffic. This is what proves “sourced cheapest” — Atlas mode draws on these per call to serve the cheapest path that holds quality.

ProviderQuantInput $/1MOutput $/1MLatency p50 / p95Throughput p50 / p95Uptime 30mSuccess
DeepInfraq4· fp4$0.5000$2.1500537ms / 1365ms54 / 76 tok/s100.00%100.0%
SiliconFlowq8· fp8$0.5000$2.18001706ms / 5013ms18 / 21 tok/s100.00%99.8%
AtlasCloudq8· fp8$0.5500$2.15002794ms / 6419ms14 / 20 tok/s99.53%99.8%
StreamLakeundisclosed$0.5710$2.28602610ms / 6994ms22.5 / 26 tok/s100.00%100.0%
Novitaq8· fp8$0.7000$2.50002138ms / 8657ms23 / 28.5 tok/s100.0%

“—” means live telemetry hasn’t accumulated enough recent traffic for that endpoint. “undisclosed” means the provider serves the model but doesn’t expose the quantization label (typically running fp8 / int8 internally).

Intelligence breakdown

Composite score is a weighted average of public benchmarks (30% MMLU-Pro, 25% code pass@1, 25% math, 20% GPQA). Numbers come from model cards and the Artificial Analysis intelligence harness; missing components are renormalised over what’s present.

MMLU-Pro

85.0

broad reasoning

Code

73.3

pass@1 (HumanEval / LiveCodeBench)

AIME 2025

87.5

math accuracy

GPQA Diamond

81.0

hard reasoning

Source: DeepSeek-R1-0528 model card (MMLU-Pro 85.0, GPQA-Diamond 81.0, LiveCodeBench 73.3, AIME25 87.5); huggingface.co/deepseek-ai/DeepSeek-R1-0528

How Atlas mode sources DeepSeek: R1 0528

  • Strict mode — pin DeepSeek: R1 0528 exactly and we pass it straight through, sourced from the cheapest provider above. The same model, no substitutions — currently DeepInfra at $0.50/1M.
  • Atlas mode — the default. Each call is auto-optimized for the cheapest path that holds quality, at least 5% off going direct from call one and climbing as it ramps. You always see which model served the call and exactly what you saved — thumbs-down anything you don’t like for a full refund.