PRICE~moonshotai/kimi-latest cached_input_per_mtok increased from 0.29 to 0.8018mPRICE~moonshotai/kimi-latest output_per_mtok increased from 11.36 to 13.0018mPRICE~moonshotai/kimi-latest input_per_mtok increased from 0.75 to 0.9918mPERFORMANCEGLM-5.2 → Scaleway: TTFT ↑ 20%48mPERFORMANCEDeepSeek V4 Flash → Scaleway: TTFT ↑ 35%48mPERFORMANCEMistral Medium 3.5 → Scaleway: TTFT ↓ 21%48mPERFORMANCEDeepSeek V4 Flash → Scaleway: throughput ↓ 20%48mPERFORMANCEGPT-OSS 120B → Scaleway: throughput ↑ 31%48mPERFORMANCEGPT-OSS 120B → Scaleway: TTFT ↓ 24%48mPERFORMANCEGPT-OSS 120B → Together AI: throughput ↑ 20%48mPERFORMANCEKimi K3 via OpenRouter: TTFT ↓ 49%48mPERFORMANCEGLM-5.2 via OpenRouter: reliability recovered 96.8% → 100.0%48mPERFORMANCEKimi K3 via Cortecs: throughput ↑ 184%48mPERFORMANCEKimi K3 via Cortecs: TTFT ↓ 42%48mPERFORMANCEGLM-5.2 via OpenRouter: throughput ↓ 49%48mPERFORMANCEGLM-5.2 via OpenRouter: TTFT ↑ 121%48mPERFORMANCEMiniMax M3 via OpenRouter: reliability recovered 96.8% → 100.0%48mPERFORMANCEDeepSeek V4 Pro via Cortecs: throughput ↑ 27%48mPERFORMANCEGPT-OSS 120B via OpenRouter: throughput ↓ 22%48mPERFORMANCEGLM-4.7 via OpenRouter: TTFT ↓ 78%48mPERFORMANCEGPT-OSS 20B → Groq: TTFT ↑ 24%48mPERFORMANCEKimi K3 → Together AI: TTFT ↓ 34%48mPERFORMANCEDeepSeek V4 Flash via OpenRouter: throughput ↓ 17%48mPERFORMANCELlama 3.3 70B via Cortecs: TTFT ↓ 51%48mPERFORMANCEQwen3 235B via OpenRouter: throughput ↑ 37%48mPERFORMANCEQwen3 235B via OpenRouter: TTFT ↓ 29%48mPERFORMANCEDeepSeek V4 Flash → Together AI: throughput ↑ 19%48mPERFORMANCEGLM-5.3 via OpenRouter: throughput ↓ 17%48mPERFORMANCEGPT-OSS 20B via Cortecs: throughput ↑ 28%48mPERFORMANCELlama 3.3 70B → Together AI: TTFT ↓ 33%48mPRICE~moonshotai/kimi-latest cached_input_per_mtok decreased from 0.80 to 0.2949mPRICE~moonshotai/kimi-latest output_per_mtok decreased from 13.00 to 11.3649mPRICE~moonshotai/kimi-latest input_per_mtok decreased from 0.99 to 0.7549mPRICE~deepseek/deepseek-v4-flash-latest cached_input_per_mtok decreased from 0.0077 to 0.001149mPRICE~deepseek/deepseek-v4-flash-latest output_per_mtok decreased from 1.28 to 1.0449mPRICE~deepseek/deepseek-v4-flash-latest input_per_mtok decreased from 0.0077 to 0.003849mPRICEz-ai/glm-5.1 cached_input_per_mtok increased from 0.18 to 0.2649mPRICEz-ai/glm-5.1 output_per_mtok increased from 3.03 to 4.4049mPRICEz-ai/glm-5.1 input_per_mtok increased from 0.96 to 1.4049mPRICEtencent/hy4-preview cached_input_per_mtok increased from 0.038 to 0.04249m

Cerebras vs OpenRouter

3 shared models · US vs US · independent benchmarks

Registry facts

FactCerebrasOpenRouter
Typeinference hostrouter
HeadquartersUSUS
Catalogued endpoints11
EU residencyNoNo

GPT-OSS 120B: Cerebras vs OpenRouter

● MEASURED Cerebras detail → · ● MEASURED OpenRouter detail →

MetricCerebrasOpenRouter
TTFT p50184 ms474 ms
TTFT p95323 ms1413 ms
Throughput1749 tok/s68 tok/s
Reliability100%99.1%
Tool calling100%100%
Structured output100%100%
$ output / M tokens$0.75$0.17

Every figure above is an independent measurement from the canonical vantage; unmeasured fields show — until measured. How we benchmark →