frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?

https://minimaxir.com/2026/10/cheap-llms/
20•minimaxir•1h ago

Comments

verdverm•52m ago
I fully expect token vendors to go the way of mobile data providers. Increasing throughput and efficiency, yet commoditized and not reaping the massive rewards of the economic gains their technologies enable.

Can't wait for the unlimited plans we will have in due time.

Iolaum•49m ago
They will not come. You can only spend so much time speaking on the phone.

But if you get unlimited tokens, with scheduled/automated jobs and subagents you will burn a LOT of money.

verdverm•46m ago
Mobile video data and tethered devices would tell a different story. Usage growth has been significant.

I expect they will be "unlimited" in the same sense the mobile / ISPs do that 99% of people never reach the limits.

We could have "unlimited" electricity too, if we choose, by printing solar panels (like the Chinese) and hopefully fusion before long.

ModernMech•40m ago
> You can only spend so much time speaking on the phone.

Tell that to my dial up modem as I played Active Worlds in 1996.

boredatoms•29m ago
Yeu were still limited to about 4GiB/month usage as a time/bandwidth ceiling
Skunkleton•34m ago
"Unlimited" for a consumer will probably take a form like limited parallel agents at max intelligence for X amount of tokens, followed by either single threaded agents or nerfed intelligence or both.

This is the same mechanism used in unlimited mobile internet for example.

cicko•13m ago
I rarely use more than 25% of my subscription's weekly allowance. So, it's already unlimited.
JumpCrisscross•32m ago
> if you get unlimited tokens, with scheduled/automated jobs and subagents you will burn a LOT of money

Maybe, maybe not. There could be a saturation point of compute demand from ordinary consumers. (See: hot-water analogy.)

bikvfeee•30m ago
> Can't wait for the unlimited plans we will have in due time.

I don't see how this can happen. The cost of serving a model grows linearly with use (energy). The cost of serving a mobile data plan is approximately zero.

buckleyourshoe•38m ago
Competition drives prices down. We tried Kimi and GLM models after hitting 'cybersecurity' blocks on OpenAI and Anthropic. We found them to be much cheaper to access and helped us complete our work without 'cybersecurity' refusals blocking us.
soupfordummies•12m ago
I wouldn’t be surprised if US “bans” access to the Chinese models by the end of the year
bethekidyouwant•10m ago
How?
ApolloFortyNine•5m ago
If you are a legally operating company in the US just saying it's banned is enough, the threat of enforcement is enough pressure.

You and I obviously can just use a VPN, but 99% of usage is likely as part of a company.

rcstank•5m ago
What harness are you all using? OpenCode?
gum_wobble•37m ago
Because they have found a way that cost them less to have you use a lesser capable model o less expensive
hauntingseaweed•28m ago
Price war. Duh. There is no moat. Also them chinese competition. It’s going to be race to the bottom. It’s gonna be ugly for the stock market. Good for consumers though!
SugarReflex•25m ago
Can they do the same thing with fuel prices?
dwattttt•22m ago
If you make it easy for lots of people to make the fuel, yes. EVs and smaller scale power generation anyone?
YetAnotherNick•22m ago
Yes if China somehow gets the fuel and subsidize it to sell it for cheaper in the global world, even the gulf state price will go down.
dymk•16m ago
Solar panels (which continue get cheaper every year) on my house makes charging my low-TCO EV free in 5 years
3eb7988a1663•24m ago
Just wait until the government ban on the dangerous open models materializes.
Kuyawa•12m ago
Duh, China?
bhu8•10m ago
It’s probably more efficient caching methods and smaller models (6 sol is actually 6 terra and opus 5.5 is smaller than 5.1 etc). Don’t overthink this please
chakintosh•9m ago
It’s a race to the bottom to see who blinks first and goes bankrupt. It’s not sustainable.

Stop resetting privacy settings to default on updates #17107

https://github.com/anthropics/claude-code/issues/17107
1•DesaiAshu•23s ago•0 comments

Show HN: Glitchrupt: Glitch your desktop when your AI agent needs your attention

https://github.com/elbee19/glitchrupt
1•impish19•28s ago•0 comments

I want my Wednesday back

https://blog.luitjes.it/posts/i-want-my-wednesday-back/
1•birdculture•32s ago•0 comments

Germany's Idle Nuclear Plants Eye New Life as AI Data Centers

https://www.bloomberg.com/news/articles/2026-10-22/germany-s-idle-nuclear-plants-eye-new-life-as-...
1•gnabgib•2m ago•1 comments

Disadvantages of Designating Mathematicians for AI Output Releases

https://terrytao.wordpress.com/2026/10/25/disadvantages-of-designating-mathematicians-for-ai-outp...
1•smilelamp•4m ago•0 comments

Reasons to keep TF state files small (2024)

https://newsletter.masterpoint.io/p/3-reasons-to-keep-tf-state-files-small
1•mooreds•6m ago•0 comments

Nvidia eyes converting old nuclear plants into AI data centers

https://www.tomshardware.com/tech-industry/data-centers/nvidia-eyes-converting-old-nuclear-plants...
2•rbanffy•11m ago•1 comments

Show HN: Visualization of Top News Stories

https://www.topaztee.com/digglabs/
1•topaztee•12m ago•0 comments

Musk says SpaceX restricted Ukraine's Starlink connection as service restored

https://www.bbc.com/news/articles/cvx28ljp8n4xo
2•cisc•13m ago•0 comments

F45 (Fedora 45) Final Is No-Go

https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org/thread/EWDOYT6TOMFY4X...
2•gnabgib•13m ago•0 comments

Intelligence vs. cost scaled by real prices

https://www.parepdf.com/lab/openrouter-price-vs-intelligence.html
1•redman25•15m ago•1 comments

Fedora 45 Hits a Second No-Go: Not Releasing Until November

https://www.phoronix.com/news/Fedora-45-Second-NO-GO
1•rbanffy•19m ago•1 comments

Europe's summer heat wave killed more han 50k people in urban areas

https://www.science.org/content/article/europe-s-summer-heat-wave-killed-more-50-000-people-urban...
2•rbanffy•20m ago•0 comments

Black October (Information Security)

https://en.wikipedia.org/wiki/Black_October_(information_security)
1•tech234a•21m ago•0 comments

Show HN: Valence – Self-hosted media browser (Jellyfin/Plex alternative)

https://www.getvalence.app/
1•MarquesCoding•25m ago•0 comments

Show HN: I built an AI agent to find customers on Telegram while I sleep

https://opensparta.com/en
2•hp2090•29m ago•0 comments

A rising tide raises all rents

https://jyn.dev/rising-tide-raises-rents/
3•iamnothere•32m ago•0 comments

Show HN: A Blender addon that makes Claude twice as efficient at using nodes

https://github.com/carson-katri/geometry-script
1•carson-katri•32m ago•1 comments

Abstraction on Abstraction on Abstraction

1•cccxccccxcccxxx•33m ago•0 comments

Odyssey series made with Grok Imagine

https://twitter.com/imagine/status/2113712964419063911
1•kklisura•33m ago•0 comments

Show HN: Jet Pilot – An open-source, native Kubernetes desktop client

https://github.com/unxsist/jet-pilot
2•unxsist•34m ago•0 comments

TryMyRepo

https://www.trymyrepo.com/
1•indigodaddy•37m ago•1 comments

What Security Teams Need to Know About Agents Using Abliterated Models

https://eve.security/blog/what-security-teams-need-to-know-about-agents-using-abliterated-models
2•mooreds•37m ago•0 comments

Pi on a Local Model

https://billdx.github.io/Pi-Qwen/
1•mooreds•38m ago•0 comments

Worried by Wordexp(3)

https://www.cipht.net/2023/11/21/worried-by-wordexp.html
1•fanf2•40m ago•1 comments

A Strongly Worded Letter: Mistakes Claude Makes with Hardware in the Loop

https://groundrun.io/blog/mistakes-claude-makes-with-hardware-in-the-loop/
1•adunk•41m ago•0 comments

China increasingly depends on copying American AI

https://economist.com/science-and-technology/2026/10/25/china-increasingly-depends-on-copying-ame...
3•andsoitis•41m ago•0 comments

Vivisection of an Active Russian-Market MAAS Platform and PyMapper Distribution [pdf]

https://papers.lisdex.sh/hermes/paper.pdf
1•carterperez-dev•42m ago•0 comments

Aster My HobbyOS with a custom C toolchain and network stack (runs on 128MB RAM)

https://github.com/antoniodev84/Aster
2•antoniolornunes•45m ago•0 comments

Sliding into Silence? We Are Speaking 300 Daily Words Fewer Every Year

https://journals.sagepub.com/doi/10.1177/17456916261425131
1•rzk•46m ago•0 comments