Tokens per Second | Kyle JeongⅡPause↻Reset

1,000 output tokens<br>One trip to the other side = 1,000 output tokens · A round trip = 2,000

gpt-oss-120B (low) Cerebras · 1,538.0 t/s<br>120B params ↗

gpt-oss-20B (high) Groq · 949.5 t/s<br>20B params ↗

GPT-5.6 Sol (ultrafast preview) Cerebras · up to 750 t/s<br>Undisclosed params ↗

Nemotron 3.5 Lightning DeepInfra (NVFP4) · 516.2 t/s<br>31.6B / 3.6B params ↗

Gemini 3.7 Flash (high) Google AI Studio · 340.1 t/s<br>Undisclosed params ↗

DeepSeek V4 Flash (Non-reasoning) Makora · 274.1 t/s<br>284B / 13B params ↗

MiniMax-M3 Modular · 231.5 t/s<br>428B / 23B params ↗

Muse Spark 1.2 (xhigh) Meta · ~217.2 t/s<br>Undisclosed params ↗

GLM-5.2 (max) Baseten (FAST) · 195.5 t/s<br>744B params ↗

Gemini 3.5 Flash (high) Google AI Studio · 172.9 t/s<br>Undisclosed params ↗

GPT-5.6 Luna (max) OpenAI · 150.2 t/s<br>Undisclosed params ↗

Claude Sonnet 5 (Adaptive Reasoning, Max Effort) Google · 81.8 t/s<br>Undisclosed params ↗

Grok 4.6 (high) SpaceXAI · 65.5 t/s<br>Undisclosed params ↗

Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) Anthropic · 63.2 t/s<br>Undisclosed params ↗

GPT-5.6 Sol (max) OpenAI · 61.7 t/s<br>Undisclosed params ↗

Claude Opus 5 (Adaptive Reasoning, Max Effort) Anthropic · 54.1 t/s<br>Undisclosed params ↗

Kimi K3 (max) Fireworks · 48.9 t/s<br>2.8T params ↗

Qwen3.8 Max Alibaba Cloud · 47.0 t/s<br>Undisclosed params ↗