“And models keep getting bigger and more expensive, too.”
Our reading, the sentence that was checked:
AI models keep getting larger and more expensive.
Assessment: Unresolved Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, How To Argue With An AI Booster, paragraph 243.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-how-to-argue-with-an-ai-booster-a265e4-a1
We recorded 35 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-08-25. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- models keep getting bigger
- more expensive
Our interpretation
- The plain reading is an ongoing trend across leading general-purpose models, rather than a claim that every release is larger and dearer.
- Bigger ordinarily means parameter count; total and active parameters must be distinguished.
- Because the surrounding passage discusses inference and explicitly rejects per-token comparisons alone, more expensive is read as higher cost to run comparable application tasks.
- Use the release cycle from September 2024 through the post date: the context identifies September 2024 as the beginning of the reasoning-model shift. This is an inferred observation window, not a period explicitly attached to this sentence.
Counts as supported if: Representative successive releases in the target period show rising model size and rising inference cost on comparable tasks, including reasoning tokens.
Counts as contradicted if: Comparable evidence across representative releases shows that either model size or task-level inference cost is flat or falling, contrary to the asserted general trend.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The evidence does not establish the joint trend as stated. Qwen reports comparable capability with fewer active parameters, while Stanford documents falling per-token prices at specified capability levels. Neither result settles the size and full task cost of representative successive models during the reasoning-model shift. Calling the claim true would require missing comparisons; calling it false solely from smaller models or cheaper tokens would also overreach.
Limitations
- The observation window and cost measure require interpretation.
- One model family cannot establish a market-wide trend.
- Per-token price declines are not equivalent to lower application bills.
Evidence (2 items)
Context only: Qwen Team,
they achieve similar performance to Qwen2.5 dense base models while using only 10% of the active parameters.
Why it matters here: This is a concrete counterexample to equating capability gains with larger active models. It concerns one family and active parameters, so it does not establish the market-wide trend in total size or application cost.
- Publisher
- Qwen Team
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Context only: Stanford HAI,
LLM inference prices have fallen anywhere from 9 to 900 times per year.
Why it matters here: The section measures prices per million tokens at benchmark capability levels, with its stated comparison ending in October 2024. It cautions against a blanket cost-growth claim but does not measure the later reasoning workloads described by Zitron.
- Publisher
- Stanford HAI
- Kind of source
- Independent measurement
- Describes events of
- Public from
What would settle it
A historical comparison across leading developers combining total and active parameter counts with the cost of completing the same workloads, including reasoning tokens.
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, How To Argue With An AI Booster
And models keep getting bigger and more expensive, too.
- Where
- paragraph 243
- Published
- 2025-08-25
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-how-to-argue-with-an-ai-booster-a265e4-a1: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
About assessment wyea-how-to-argue-with-an-ai-booster-a265e4-a1.