BENCHMARKFirst benchmarks recorded: MiniMax M3 → Cortecs → ?4hBENCHMARKFirst 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

Llama 3.3 70B

10 providers · 70B dense · 128k context · 2 regions · all models →

Llama 3.3 70B across 10 providers

Same model. Different infrastructure. DEV DATA
TTFT spread3.4×Throughput spread28.4×Output price3.3×Tool reliability4.7 pts
TTFT110 ms379 msbest: Cerebras
Throughput100 tok/s2835 tok/sbest: Cerebras
Output price$0.20/M$0.66/Mbest: DeepInfra
Reliability96.9%99.0%best: Fireworks
Tool success94.5%99.2%best: OpenRouter

Llama 3.3 70B — tool-calling benchmark

10 providers · 94.5% → 99.2% · 4.7 pt spread · DEV DATA
#ProviderToolsΔ bestRel.JSONTTFTP95Tok/s$ Out
01OpenRouter99.2%—98.0%98.6%284 ms973 ms216$0.48
02Fireworks98.8%−0.499.0%99.9%231 ms790 ms251$0.48
03Cerebras98.2%−1.098.9%98.5%110 ms272 ms2835$0.66
04Groq98.1%−1.198.5%98.6%132 ms298 ms1013$0.32
05Nebius98.0%−1.298.8%99.3%213 ms778 ms192$0.22
06Together97.5%−1.797.7%97.8%297 ms916 ms176$0.40
07Nextbit96.7%−2.597.8%96.9%252 ms890 ms113$0.24
08Scaleway96.4%−2.897.9%97.7%263 ms973 ms130$0.40
09DeepInfra95.5%−3.796.9%98.6%379 ms1212 ms119$0.20
10OVHcloud94.5%−4.797.5%96.4%275 ms703 ms100$0.37