BENCHMARKFirst benchmarks recorded: MiniMax M3 → Cortecs → ?3hBENCHMARKFirst benchmarks recorded: Kimi K3 → Cortecs → ?4hBENCHMARKFirst benchmarks recorded: DeepSeek V4 Pro → Cortecs → ?4hBENCHMARKFirst benchmarks recorded: GLM-4.7 → Cortecs → ?4hBENCHMARKFirst benchmarks recorded: GLM-5.2 → Cortecs → ?4h
BENCHMARK

DeepSeek V3 Across Providers

One open model, eight infrastructures, materially different results.

Aug 15, 2026103,921 benchmark runs13 providers8 modelsMethodology v0.10-devDEV DATA
Key findings
  • Nebius runs DeepSeek-V3 best (94.5) among 6 providers.
  • TTFT spans 340–693 ms across the pool for identical weights.
  • The cheapest serving is $0.39/M output — 2.8× below the most expensive.

The provider ranking

One row per provider, scored under the General profile. The full interactive table, with every metric column, lives in the product.

#ProviderScoreTTFTTok/s$ Out
01Nebius94.5340 ms116$0.50
02Fireworks92.6368 ms153$1.10
03Together76.9473 ms107$0.90
04OpenRouter75.3453 ms131$1.10
05DeepInfra69.5604 ms72$0.45
06SiliconFlow61.2693 ms75$0.39
DeepSeek-V3 provider ranking · DEV DATA

The ranges

The per-metric spread for this one model demonstrates the site's thesis in miniature.

TTFT340 ms693 ms
Throughput72 tok/s153 tok/s
Output price$0.39/M$1.10/M
Reliability96.6%99.0%
DeepSeek-V3 across providers — per-metric ranges · DEV DATA
What this means for engineers
  • DeepSeek-V3 deployments should be provider-benchmarked before launch: the winner differs by workload profile.
Dataset
103,921 runs30-day rolling window · Aug 2026snapshot 2026-08-15-v1Methodology v0.10-devDEV DATA

Published figures are pinned to this snapshot and are not recalculated retrospectively. View the current benchmark →

Cite this researchInferenceBench (2026). DeepSeek V3 Across Providers. InferenceBench Research.
← All research
Related research
MARKET REPORT

The State of AI Inference — August 2026

The flagship snapshot of the inference market: the global provider ranking, the price–performance frontier, and the growing spread between infrastructures serving identical models.

Aug 15, 2026MARKET REPORT

The European AI Inference Report

The deepest public look at European inference: EU-owned providers versus Europe-hosted endpoints, the residency premium, and where the region genuinely competes.

Aug 15, 2026ENGINEERING STUDY

Same Model, Different Infrastructure: How Large Is the Gap?

The core InferenceBench thesis, measured: for the same model, TTFT varies by multiples, throughput by an order of magnitude, and tool reliability by whole points.

Aug 15, 2026