Side by side · same API

Compare models

Price, speed and capacity for two models on one page. Both take the same request on /v1/chat/completions, so switching between them is a change to one string.

Common pairs

Comparing

Change either model and the link updates, so a comparison can be shared.

Model AServerless

by Meta · FP8 · 128K context

chat
dedicated

Model page

Model BServerless

by Alibaba · FP8 · 128K context

chat
reasoning
dedicated

Model page

Price · serverless, USD per 1M tokens

Input$0.54$0.3044% less

Output$0.72$0.5031% less

Blended, half in and half out$0.63$0.4037% less

Speed · median, last 24 hours

First token, p50212 ms150 ms62 ms faster

Decode speed, one request96 tok/s130 tok/s35% faster

Capacity

Served context128K tokens128K tokens

QuantizationFP8FP8

Taskschatchat · reasoning

Regionsus-east · us-west · eu-westus-east · us-west · eu-west

Dedicated · one replica

Replica2× H100 · $4.98 / hour1× H100 · $2.49 / hourhalf the rate

Peak throughput6,000 tok/s20% more5,000 tok/s

Break-even vs serverless5.77B tok/mo4.54B tok/mo

A highlighted difference marks the better number in a row. No mark means a tie, or nothing to rank.

Estimate

At your volume

Serverless price, with half input and half output tokens, the same split the calculator uses.

Monthly tokens

Llama 3.3 70B$630 / month
Qwen3 32B$400 / month

Qwen3 32B costs $230 less a month at 1B tokens.

Open the calculator

API

Same request, one string changed

Both models take the same body on /v1/chat/completions. Moving from one to the other is a change to the model field.

diff1 line changed

 curl https://api.weir.example/v1/chat/completions \
   -H "Authorization: Bearer $WEIR_API_KEY" \
   -H "Content-Type: application/json" \
   -d '{
-Removed line:     "model": "llama-3-3-70b",
+Added line:     "model": "qwen3-32b",
     "messages": [
       { "role": "user", "content": "What does a KV cache do?" }
     ]
   }'

Get an API key Read the quickstart