It seems to me that large software companies often adopt an open-source approach initially to attract enthusiasts and stand out from leading competitors, but tend to adopt similar philosophies as their rivals once they achieve significant recognition or investment.
You can see this as an endgame of 'commoditize your complement' (https://gwern.net/complement): you're happy to contribute to commodification while you are the small scrappy player, but at some point, if you are really successful, you will want to pull up the ladder after yourself, as it were.
Idle question: what's the complement of the model? The GPUs? By that logic, the hardware companies will be your best bet for open model development. (Everything old is new again).
The complement play for Mistral was that they'd be the low-cost SaaS serving their own models or any better ones that the FLOSS releases created and the upsell: you prototype and develop on Mistral models like Mixtral and then when you become a big boy, you outsource hosting or customization to them (with their razor-thin margins enforced by competition). The model releases generate demand for their low-cost hosting and consulting. (Think Red Hat, not OpenAI.)
As opposed to an all-inclusive API where the pricing can be as high as the consumer-surplus with very fat margins, because it's all-or-nothing.
Comments
It seems to me that large software companies often adopt an open-source approach initially to attract enthusiasts and stand out from leading competitors, but tend to adopt similar philosophies as their rivals once they achieve significant recognition or investment.
It's all a marketing tactic, and I ain't for it.
You can see this as an endgame of 'commoditize your complement' (https://gwern.net/complement): you're happy to contribute to commodification while you are the small scrappy player, but at some point, if you are really successful, you will want to pull up the ladder after yourself, as it were.
Idle question: what's the complement of the model? The GPUs? By that logic, the hardware companies will be your best bet for open model development. (Everything old is new again).
The complement play for Mistral was that they'd be the low-cost SaaS serving their own models or any better ones that the FLOSS releases created and the upsell: you prototype and develop on Mistral models like Mixtral and then when you become a big boy, you outsource hosting or customization to them (with their razor-thin margins enforced by competition). The model releases generate demand for their low-cost hosting and consulting. (Think Red Hat, not OpenAI.)
As opposed to an all-inclusive API where the pricing can be as high as the consumer-surplus with very fat margins, because it's all-or-nothing.