Back to home
Speed
How fast can each AI model generate output? Ranked by tokens per second throughput.
RankModelScore
12345678
9
10
111213141516
17181920
2122232425262728
2930
I
Mercury 2inception/mercury-2
1036.6 t/s
I
Granite 3.3 8Bibm/granite-3-3-8b-instruct
420.0 t/s
I
Granite 4.0 H Smallibm/granite-4-0-h-small
410.4 t/s
L
LFM2.5 VL 1.6Bliquidai/lfm2-5-vl-1-6b
390.9 t/s
S
Step 3.7 Flashstepfun/step-3-7-flash
385.2 t/s
M
HyperNova 60B 2605multiversecomputing/hypernova-60b
344.0 t/s
L
LFM2.5 8B A1Bliquidai/lfm2-5-8b-a1b
343.1 t/s
Nemotron 3 Nano Omni 30B A3B Rnvidia/nemotron-3-nano-omni-30b-a3b
328.1 t/s
gpt oss 120bopenai/gpt-oss-120b-low
313.8 t/s
Gemini 3.1 Flash Litegoogle/gemini-3-1-flash-lite-preview
295.4 t/s
NVIDIA Nemotron Nano 12B v2 VLnvidia/nvidia-nemotron-nano-12b-v2-vl-reasoning
280.9 t/s
Llama 3.1 Nemotron Instruct 70nvidia/llama-3-1-nemotron-instruct-70b
280.3 t/s
S
Step 3.5 Flashstepfun/step-3-5-flash-0202
277.1 t/s
S
Step 3.5 Flash 2603stepfun/step-3-5-flash
262.9 t/s
gpt oss 20bopenai/gpt-oss-20b-low
260.5 t/s
Nova Microaws/nova-micro
250.4 t/s
Gemini 2.5 Flash Litegoogle/gemini-2-5-flash-lite-reasoning
239.1 t/s
o3openai/o3-mini-high
237.6 t/s
Gemini 3.5 Flashgoogle/gemini-3-5-flash
236.9 t/s
Qwen3.5 Omni Flashalibaba/qwen3-5-omni-flash
230.7 t/s
Gemini 2.5 Flashgoogle/gemini-2-5-flash-reasoning
213.9 t/s
S
Sarvam 30Bsarvam/sarvam-30b
209.3 t/s
GPT-5openai/gpt-5-1-codex-mini
208.0 t/s
Nova 2.0 Liteaws/nova-2-0-lite
203.6 t/s
Qwen3.7 Maxalibaba/qwen3-7-max
200.8 t/s
Gemini 3 Flashgoogle/gemini-3-flash
199.7 t/s
GPT-5openai/gpt-5-6-luna
197.1 t/s
Command A+cohere/command-a-plus
194.3 t/s
GPT-4oopenai/gpt-4o
187.4 t/s
Nemotron 3 Ultra 550B A55Bnvidia/nvidia-nemotron-3-ultra-550b-a55b
186.4 t/s
394 models tested