NEWMEN
← back to all models

DeepSeek: DeepSeek V4 Pro

deepseek/deepseek-v4-pro

DeepSeekopen-weight1.0M context14 providersIntelligence 90.2

Cheapest provider

$0.43 / 1M

DeepSeek

Fastest provider (p95)

No throughput data yet — populated as traffic accumulates

Intelligence (composite)

90.2

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
DeepSeekundisclosed$0.4350$0.870099.99%
Baiduq8· fp8$0.7605$1.521098.92%
DeepInfraq4· fp4$1.3000$2.600099.45%
GMICloudq8· fp8$1.3920$2.784098.26%
StreamLakeundisclosed$1.5660$3.132098.58%
Novitaq8· fp8$1.6000$3.200099.70%
SiliconFlowq8· fp8$1.6000$3.480099.55%
Alibabaundisclosed$1.6080$3.216097.53%
AtlasCloudq8· fp8$1.6800$3.380097.86%
Veniceundisclosed$1.7300$3.796096.74%
DigitalOceanundisclosed$1.7400$3.480097.89%
Fireworksundisclosed$1.7400$3.480096.87%
Parasailq8· fp8$1.7400$3.480096.12%
Togetherundisclosed$2.1000$4.400091.42%

“—” 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

87.5

broad reasoning

Code

93.5

pass@1 (HumanEval / LiveCodeBench)

MATH

math accuracy

GPQA Diamond

90.1

hard reasoning

Source: DeepSeek V4-Pro Max (MMLU-Pro 87.5, GPQA-Diamond 90.1, LiveCodeBench 93.5); framia / morphllm benchmark tables + Artificial Analysis Intelligence Index ~52

How Atlas mode sources DeepSeek: DeepSeek V4 Pro

  • Strict mode — pin DeepSeek: DeepSeek V4 Pro exactly and we pass it straight through, sourced from the cheapest provider above. The same model, no substitutions — currently DeepSeek at $0.43/1M.
  • 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.