jpschroederinc/artificial13d agoNo talks about how cheap open source AI models are anymore.@OpenAI has entirely crushed that conversation with Luna.#ai#llms#openai#open source010 comments
capajjx13d ago@jpschroeder @OpenAI nope because the numbers on @ArtificialAnlys for glm-5.3-flash are wildly different. https://openrouter.ai/compare/openai/gpt-6-luna/z-ai/glm-5.3-flash real inference cost is like 5x less so in fact it's @Zai_org crushing everyone😄0 Replyjpschroederx13d agois your argument that their price per token is less?0 Replycapajjx13d agoThird party providers have the model for price 80 percent lower per million tokens0 Replyjpschroederx13d agoAnd yet it's more expensive per task.0 Replycapajjx13d agoNo it's not. This is measured with the official pricing not the third party pricing0 Replybulquerque1x13d agoprice per token * tokens per task = price per task Glm 5.3 flash takes a lot more tokens0 Replycapajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Replyyornfiftyx12d agoyou read it wrong then, this is the only price you should look into glm cost 3 times more0 Reply
jpschroederx13d agois your argument that their price per token is less?0 Replycapajjx13d agoThird party providers have the model for price 80 percent lower per million tokens0 Replyjpschroederx13d agoAnd yet it's more expensive per task.0 Replycapajjx13d agoNo it's not. This is measured with the official pricing not the third party pricing0 Replybulquerque1x13d agoprice per token * tokens per task = price per task Glm 5.3 flash takes a lot more tokens0 Replycapajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
capajjx13d agoThird party providers have the model for price 80 percent lower per million tokens0 Replyjpschroederx13d agoAnd yet it's more expensive per task.0 Replycapajjx13d agoNo it's not. This is measured with the official pricing not the third party pricing0 Replybulquerque1x13d agoprice per token * tokens per task = price per task Glm 5.3 flash takes a lot more tokens0 Replycapajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
jpschroederx13d agoAnd yet it's more expensive per task.0 Replycapajjx13d agoNo it's not. This is measured with the official pricing not the third party pricing0 Replybulquerque1x13d agoprice per token * tokens per task = price per task Glm 5.3 flash takes a lot more tokens0 Replycapajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
capajjx13d agoNo it's not. This is measured with the official pricing not the third party pricing0 Replybulquerque1x13d agoprice per token * tokens per task = price per task Glm 5.3 flash takes a lot more tokens0 Replycapajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
bulquerque1x13d agoprice per token * tokens per task = price per task Glm 5.3 flash takes a lot more tokens0 Replycapajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
capajjx13d agoMaybe but it's still cheaper than Luna at the current floor price 😔0 Replybulquerque1x12d agoLuna is almost free.0 Replyfaykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
faykn_x12d agocache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less. if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.0 Reply
yornfiftyx12d agoyou read it wrong then, this is the only price you should look into glm cost 3 times more0 Reply
@jpschroeder @OpenAI nope because the numbers on @ArtificialAnlys for glm-5.3-flash are wildly different.
https://openrouter.ai/compare/openai/gpt-6-luna/z-ai/glm-5.3-flash
real inference cost is like 5x less so in fact it's @Zai_org crushing everyone😄
is your argument that their price per token is less?
Third party providers have the model for price 80 percent lower per million tokens
And yet it's more expensive per task.
No it's not. This is measured with the official pricing not the third party pricing
price per token * tokens per task = price per task
Glm 5.3 flash takes a lot more tokens
Maybe but it's still cheaper than Luna at the current floor price 😔
Luna is almost free.
cache read is over half the price (most of the tokens used in agentic work are cached input) and per task tokens used are also way less.
if you need a cheap as shit model for basic work im sure xhigh will be good enough (its only 10% dumber) and laughably cheap.
you read it wrong then, this is the only price you should look into
glm cost 3 times more