PRICE~z-ai/glm-latest cached_input_per_mtok increased from 0.08 to 0.1516mPRICE~z-ai/glm-latest output_per_mtok decreased from 4.00 to 2.7916mPRICE~z-ai/glm-latest input_per_mtok decreased from 0.12 to 0.06616mPRICE~moonshotai/kimi-latest cached_input_per_mtok decreased from 0.80 to 0.4516mPRICE~moonshotai/kimi-latest input_per_mtok decreased from 0.98 to 0.9716mPRICE~deepseek/deepseek-v4-flash-latest output_per_mtok decreased from 1.04 to 0.8116mPRICE~deepseek/deepseek-v4-flash-latest input_per_mtok increased from 0.0038 to 0.008616mPRICE~deepseek/deepseek-pro-latest cached_input_per_mtok decreased from 0.10 to 0.004216mPRICE~deepseek/deepseek-pro-latest output_per_mtok decreased from 4.20 to 0.4016mPRICE~deepseek/deepseek-pro-latest input_per_mtok increased from 0.13 to 0.1316mPRICEdeepseek/deepseek-v4-flash-0731 cached_input_per_mtok increased from 0.0051 to 0.01216mPRICEdeepseek/deepseek-v4-flash-0731 input_per_mtok increased from 0.0051 to 0.01216mPRICE~moonshotai/kimi-latest cached_input_per_mtok increased from 0.29 to 0.8047mPRICE~moonshotai/kimi-latest output_per_mtok increased from 11.36 to 13.0047mPRICE~moonshotai/kimi-latest input_per_mtok increased from 0.75 to 0.9847mPRICE~deepseek/deepseek-pro-latest input_per_mtok decreased from 0.13 to 0.131hPRICE~deepseek/deepseek-pro-latest output_per_mtok increased from 0.40 to 4.201hPRICE~deepseek/deepseek-pro-latest cached_input_per_mtok increased from 0.0042 to 0.101hPRICE~moonshotai/kimi-latest input_per_mtok decreased from 0.99 to 0.751hPRICE~moonshotai/kimi-latest output_per_mtok decreased from 13.00 to 11.361hPRICE~moonshotai/kimi-latest cached_input_per_mtok decreased from 0.80 to 0.291hPRICE~moonshotai/kimi-latest input_per_mtok increased from 0.75 to 0.992hPRICE~moonshotai/kimi-latest output_per_mtok increased from 11.36 to 13.002hPRICE~moonshotai/kimi-latest cached_input_per_mtok increased from 0.29 to 0.802hPERFORMANCEGLM-5.2 → Scaleway: TTFT ↑ 20%2hPERFORMANCEDeepSeek V4 Flash → Scaleway: TTFT ↑ 35%2hPERFORMANCEMistral Medium 3.5 → Scaleway: TTFT ↓ 21%2hPERFORMANCEDeepSeek V4 Flash → Scaleway: throughput ↓ 20%2hPERFORMANCEGPT-OSS 120B → Scaleway: throughput ↑ 31%2hPERFORMANCEGPT-OSS 120B → Scaleway: TTFT ↓ 24%2hPERFORMANCEGPT-OSS 120B → Together AI: throughput ↑ 20%2hPERFORMANCEKimi K3 via OpenRouter: TTFT ↓ 49%2hPERFORMANCEGLM-5.2 via OpenRouter: reliability recovered 96.8% → 100.0%2hPERFORMANCEKimi K3 via Cortecs: throughput ↑ 184%2hPERFORMANCEKimi K3 via Cortecs: TTFT ↓ 42%2hPERFORMANCEGLM-5.2 via OpenRouter: throughput ↓ 49%2hPERFORMANCEGLM-5.2 via OpenRouter: TTFT ↑ 121%2hPERFORMANCEMiniMax M3 via OpenRouter: reliability recovered 96.8% → 100.0%2hPERFORMANCEDeepSeek V4 Pro via Cortecs: throughput ↑ 27%2hPERFORMANCEGPT-OSS 120B via OpenRouter: throughput ↓ 22%2h

Llama 3.3 70B

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

Llama 3.3 70B provider ranking

#1 OverallScaleway99.0FastestScaleway4.82 sBest valueOpenRouterEurope #1Scaleway99.0Best toolsScaleway100.0%Best structuredScaleway100.0%
#ProviderΔ30d
01ScalewayBEST OVERALLEU RESIDENT● MEASUREDFR · Serverless · Llama 3.3 70B100.0%100.0%100.0%1446494.8s86P50↓ 0.3
02Together● MEASUREDUS · Serverless · Llama 3.3 70B100.0%98.0%100.0%61818016.9s63P25↑ 7.4
03CortecsEU RESIDENTEU OWNED● MEASUREDAT · Routed · Llama 3.3 70B100.0%98.0%100.0%44110518.3s51P75↑ 38.1
04OpenRouterBEST VALUE● MEASUREDUS · Routed · Llama 3.3 70B100.0%100.0%100.0%576503011.0s38P100↑ 1.3
4 of 7 providers ranked · · Methodology →
01Scaleway99.0
Best overallFR · Serverless · Llama 3.3 70B
TTFT144 ms
Rel.100.0%
Tok/s86
$ Out$1.03
Tools100.0%
Δ30d-0.3
View dossier
02Together94.9
US · Serverless · Llama 3.3 70B
TTFT618 ms
Rel.100.0%
Tok/s63
$ Out$1.04
Tools98.0%
Δ30d+7.4
View dossier
03Cortecs91.7
AT · Routed · Llama 3.3 70B
TTFT441 ms
Rel.100.0%
Tok/s51
$ Out$0.75
Tools98.0%
Δ30d+38.1
View dossier
04OpenRouter85.0
Best valueUS · Routed · Llama 3.3 70B
TTFT576 ms
Rel.100.0%
Tok/s38
$ Out$0.32
Tools100.0%
Δ30d+1.3
View dossier
4 of 7 providers ranked · · Methodology →
Llama 3.3 70B inference benchmark: fastest & cheapest providers · InferenceBench