Tootfinder

So to follow up on this, I've caught it in action. Models, when quantized a bit, just do a bit more poorly with short contexts. Even going from f32 (as trained) to bf16 (as usually run) to q8 tends to do okay for "normal" context windows. And q4 you start feeling like "this model is a little stupid and gets stuck sometimes” (it is! It's just that it's still mostly careening about in the space of "plausible" most of the time. Not good guesswork, but still in the zone). With long contexts, the probability of parameters collapsing to zero are higher, so the more context the more likelihood you are to see brokenness.
And then at Q2 (2 bits per parameter) or Q1, the model falls apart completely. Parameters collapse to zero easily. You start seeing "all work and no play makes jack a dull boy” sorts of behavior, with intense and unscrutinized repetition, followed by a hard stop when it just stops working.
And quantization is a parameter that a model vendor can turn relatively easily. (they have to regenerate the model from the base with more quantization, but it's a data transformation on the order of running a terabyte through a straightforward and fast process, not like training).
If you have 1000 customers and enough equipment to handle the requests of 700, going from bf16 to q8 is a no-brainer. Suddenly you can handle the load and have a little spare capacity. They get worse results, probably pay the same per token (or they're on a subscription that hides the cost anyway so you are even freer to make trade-offs. There's a reason that subscription products are kinda poorly described.)
It's also possible for them to vary this across a day: use models during quieter periods? Maybe you get an instance running a bf16 quantization. If you use it during a high use period? You get a Q4 model.
Or intelligent routing is possible. No idea if anyone is doing this, but if they monitor what you send a bit, and you generally shoot for an expensive model for simple requests? They could totally substitute a highly quantized version of the model to answer the question.
There are •so many tricks• that can be pulled here. Some of them very reasonable to make, some of them treading into outright misleading or fraudulent, and it's weirdly hard to draw the line between them.

The Richmond Free Press, a 34-year-old Black-owned weekly, shuts down due to falling ad revenue as the Black press suffers from the dispersion of its readership (Scott Nover/Washington Post)

A Black newspaper in the former capital of the Confederacy is closing
The Richmond Free Press, a Black newspaper in Virginia’s capital city, ceases publication after three decades.

Opponents of VA Redistricting Amendment Send Out Mailer/Text Falsely
-- and Offensively!
Claiming That Fighting Trump’s Assault on Our Democracy Is
“Just Like Jim Crow”
Note that many of the most prominent supporters of this amendment
- Barack Obama, Louise Lucas, Don Scott, etc.
- are African American

Opponents of VA Redistricting Amendment Send Out Mailer/Text Falsely (and Offensively!) Claiming That Fighting Trump's Assault on Our Democracy Is "Just Like Jim Crow" - Blue Virginia
See below for what’s hitting Virginia’s mailboxes and smartphones (via text) today; a group called “Democracy and Justice PAC” – nothing popping up in Google about it, but the mailing address is listed as 123 Bulifants Blvd in Williamsburg, VA (same address as “Campaign Compliance Center“) – sent out the following. The problems with these […]

That time when Johnny Klimek (composer for Cloud Atlas and many others of Tom Tykwer films) and Dr. Motte (founder of the Berlin Loveparade) got together in 1996 to create their one-off project Holy Language...
Fireplace (original version, 9 minutes)
https://www.youtube.com/watch?v=DgL3DNn7OGE

Filing: Broadcom agrees to produce future versions of Google's TPUs and expands its Anthropic deal to give the startup access to ~3.5 GW of computing capacity (Jordan Novet/CNBC)
https://www.cnbc.com/2026/04/06/broadcom-agrees-to-expanded…

Broadcom agrees to expanded chip deals with Google, Anthropic
Broadcom said it agreed to produce future versions of Google artificial intelligence chips, and announced an expanded deal with Anthropic.

🇺🇦 Auf radioeins läuft...
Superspace:
🎵 Superspace Feeling (House Version)
#NowPlaying #Superspace
https://superspacemusic.bandcamp.com/track/superspace-feeling-house-version-2
https://open.spotify.com/track/4gVlkx1BUOFnjqbhUNKrNC

🇺🇦 #NowPlaying on #BBC6Music's #LaurenLaverne
Al Green:
🎵 L-O-V-E
#AlGreen
https://a-cee.bandcamp.com/track/al-green-l-o-v-e-cant-explain-this-feeling-version-a-cees-way

🇺🇦 Auf radioeins läuft...
Presley, Elvis:
🎵 You've Lost That Loving Feeling (EPiC Version)
#NowPlaying #Presley #Elvis
https://open.spotify.com/track/5zaT6hiGgLAYdOw9r5QTJa

No official version of the proposal has been made available, but a summary released by Iran's Supreme National Security Council includes demands for the following:
The Strait of Hormuz to be reopened "under the co-ordination of the armed forces of Iran".

The war against "all components" of Iran's so-called Axis of Resistance to end.

US forces to withdraw from "all bases and points of deployment within the region".

What is in Iran's 10-point peace plan?
Iranian proposal appears to include demands for reparations for war damages

🇺🇦 #NowPlaying on #BBC6Music's #RileyAndCoe
Al Green:
🎵 L-O-V-E
#AlGreen
https://a-cee.bandcamp.com/track/al-green-l-o-v-e-cant-explain-this-feeling-version-a-cees-way