Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Please don't discontinue Gemini 2.5 Flash

discuss.ai.google.dev

Post reply on HN
Print view View original

81–90 of 93 posts

Re: Please don't discontinue Gemini 2.5 Flash

#81

Post by BoorishBears »

recursive wrote: post #73
samuelknight wrote: post #23

Earlier quoted context omitted.

It's not a smell. Why should these developers rebuild a core piece of their stack every few months. Switching out a model requires a new round of testing and validation when we should be able to rely on a piece of software the behave the same way since the last time we touched it.

> Why should these developers rebuild a core piece of their stack every few months? That's what they signed up for when established a hard dependency on an subscription online-only LLM model.

If they were on open weights, at some point the provider would deprecate it, probably with worse notice

And self-hosting would probably have been more expensive unless they had massive volume: Deepseek V3.x would have been the comparable open weight model for the performance and isn't that cost effective until hosted across multiple nodes with large batch sizes

Re: Please don't discontinue Gemini 2.5 Flash

#82

Post by recursive »

BoorishBears wrote: post #81
recursive wrote: post #73

Earlier quoted context omitted.

> Why should these developers rebuild a core piece of their stack every few months? That's what they signed up for when established a hard dependency on an subscription online-only LLM model.

If they were on open weights, at some point the provider would deprecate it, probably with worse notice And self-hosting would probably have been more expensive unless they had massive volume: Deepseek V3.x would have been the comparable open weight model for the performance and isn't that cost effective until hosted across multiple nodes with large batch sizes

Sounds like something you'd need to evaluate before adopting then.

Re: Please don't discontinue Gemini 2.5 Flash

#83

Post by tiew9Vii »

segmondy wrote: post #7

This is the problem with cloud models, you build a "predictable" workflow then they remove it with a new and improved one that is less deterministic and often costs more. If you use a local model discontinuation is no longer a thing to worry about.

This is a huge (obvious) risk I've seen very little talk/concern on.

It's not even the new one being "less deterministic and often costs more".

Application prompts overfit the model they are using to get the output they want. Switch model, the prompt no longer produces the output you expect and you may not be able to get the output you need anymore.

If a model is critical for your application/business you want the ability to run it yourself otherwise you are stuck when it becomes unavailable.

Re: Please don't discontinue Gemini 2.5 Flash

#84

Post by 8n4vidtmkvmk »

captainbland wrote: post #53
svachalek wrote: post #37

Earlier quoted context omitted.

Yeah IIRC the latest Pro is $12 and Flash is $9 which is not the usual 2X-3X multiplier we see separate model grades. It also puts Flash now about 2X GLM 5.2, which is a highly capable open weight model.

I think the thing is that 3.5 flash is actually similarly capable on a lot of tasks that matter and is faster. Pro is more specialised in the direction of mathematical reasoning and stuff.

I use flash 3.5 a lot. Its significantly worse than gpt 5.5.

Re: Please don't discontinue Gemini 2.5 Flash

#85

Post by altbdoor »

whycombinetor wrote: post #36

I feel this way about gpt-5-nano (EOL December 2026). It seems like the open weight models have progressed a long way since these old models were released though. Deepseek V4 Flash is even cheaper than gpt-5-nano. I'm still going to pay a cloud provider to run it for me, I'm not local inference pilled yet, but I _can_ run it myself in the future if worse comes to worst. Objectively testable evals are one thing, but h…

[dead]

Re: Please don't discontinue Gemini 2.5 Flash

#86

Post by palmotea »

samuelknight wrote: post #23
hrpnk wrote: post #9

I love how there is a "Please do not discontinue gemini-2.0-flash[-lite], 2.5 is NOT an equivalent" from Feb 20th. Getting too attached to models is a smell.

It's not a smell. Why should these developers rebuild a core piece of their stack every few months. Switching out a model requires a new round of testing and validation when we should be able to rely on a piece of software the behave the same way since the last time we touched it.

> Why should these developers rebuild a core piece of their stack every few months. Switching out a model requires a new round of testing and validation when we should be able to rely on a piece of software the behave the same way since the last time we touched it.

Because fuck you, that's way. /s

There's a massive amount of contempt from a lot of software businesses for their customers, and it's just getting worse with LLM models.

Re: Please don't discontinue Gemini 2.5 Flash

#87

Post by Capricorn2481 »

samuelknight wrote: post #23
hrpnk wrote: post #9

I love how there is a "Please do not discontinue gemini-2.0-flash[-lite], 2.5 is NOT an equivalent" from Feb 20th. Getting too attached to models is a smell.

It's not a smell. Why should these developers rebuild a core piece of their stack every few months. Switching out a model requires a new round of testing and validation when we should be able to rely on a piece of software the behave the same way since the last time we touched it.

These are AI models designed to be non deterministic. If upgrading from GPT 5.5 > 5.6 breaks your entire workflow, you shouldn't be using AI to begin with

Re: Please don't discontinue Gemini 2.5 Flash

#89

Post by sidpatil »

segmondy wrote: post #7

This is the problem with cloud models, you build a "predictable" workflow then they remove it with a new and improved one that is less deterministic and often costs more. If you use a local model discontinuation is no longer a thing to worry about.

Further, they may change the behavior of the existing model without providing clear notice. E.g. existing models start to behave worse shortly after new model releases, due to resource reallocation away from the older models toward the newly-released ones.

Re: Please don't discontinue Gemini 2.5 Flash

#90

Post by jazzyjackson »

segmondy wrote: post #7

This is the problem with cloud models, you build a "predictable" workflow then they remove it with a new and improved one that is less deterministic and often costs more. If you use a local model discontinuation is no longer a thing to worry about.

You can run software you control in the cloud, so the cloud/local distinction isn’t really the problem/solution. Relying on proprietary models that can be discontinued is a problem, using an open weights model that can’t be taken away is a solution, whether or not those weights fit on a MacBook.
Post reply on HN