Input costs $0.36 per million tokens and output costs $0.40. It takes up to 32,768 tokens of context at once. At 10 million tokens a month that works out to about $3.70. Its intelligence index is 7.7, ranking 145 of 175 scored models. Listed in the catalogue on 9/19/2024.
The maker does not sell this model directly. The figure shown is a reseller rate on OpenRouter, and its moves are not reported as news.
You will only get news about this version (qwen-2.5-72b-instruct).
What it costs a month
Token prices translated into real usage.
- Light1M tokens a month: a personal tool or trial₩490$0.37
- Steady10M tokens a month: one small service₩4,898$3.70
- Heavy100M tokens a month: a service with real users₩48,984$37.00
Sending 10,000 Korean characters and getting 2,000 back, 1,000 times a month, costs about ₩4,212.
The same content in English costs about ₩3,240. Korean runs 1.3x.
Mixed 75% input, 25% output: a common ratio for chat usage, though real cost depends on how you use it. Exchange rate is from the last collection.
Report card
What this model is good at, by subject.
- KoreanKMMLU-HARDwe measured—not taken
- ScienceGPQA49.1#138/159
- MathAA Math Indexsaturated14#77/94
- CodingAA Coding Index—not taken
- ReadingAA-LCRin-house test—not taken
- KnowledgeMMLU-Pro72#66/92
- Instruction followingIFBench36.9#98/129
- Hard reasoningHLE3.6#157/171
Measured and published by Artificial Analysis. Each subject is a different test, so compare within a subject rather than across them.
Community
Share what using it is actually like: the part the metrics do not show.
- No notes yet. Be the first to share.
Buzz trend
How much this model was talked about over the past week.
News about this model
No news about this model yet.
Other models from this vendor
Jump straight to where the conversation is.
Compare with others
Pricing and scores side by side. Blank cells mean the price is undisclosed or the model did not take that test.