“Once these models are hosted elsewhere, and once DeepSeek's methods (which I'll get to shortly) are recreated (which won't take long), I believe we'll see that these prices are indicative of how cheap these models are to run.”
Our reading, the sentence that was checked:
Once DeepSeek's models are hosted elsewhere and its methods recreated (soon), its low prices will prove to reflect genuinely low running costs.
Assessment: Unresolved No deadline stated on 2026-10-29 Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, Deep Impact, paragraph 35.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-deep-impact-44a68b-a1
We recorded 49 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 2025-01-29. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Conditional forecast
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No deadline stated
- Deadline state on 2026-10-29
- No deadline stated (the release’s reference date)
Conditions
- once models are hosted elsewhere (antecedent, stated by the speaker)
- once DeepSeek's methods are recreated (antecedent, stated by the speaker)
From the speaker
- Once these models are hosted elsewhere
- once DeepSeek's methods ... are recreated
- which won't take long
- these prices are indicative of how cheap these models are to run
Our interpretation
- The claim concerns inference economics for the full V3 and R1 models discussed in context.
- Recreation means independently implementing the efficiency methods relevant to serving those models; it need not reproduce their original training data.
- Cheap to run means inference costs broadly compatible with the quoted API prices. It does not mean that the whole company is profitable.
- No precise deadline or required cost margin is supplied.
Counts as supported if: After external hosting and recreation of relevant methods, reproducible measurements show that both full models can serve comparable workloads at costs broadly compatible with the quoted prices.
Counts as contradicted if: After both conditions occur, comparable operating measurements establish that the quoted prices materially understate unavoidable inference costs despite the recreated methods.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Did the forecast come true?
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 deadline stated
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
External hosting and recreation of relevant inference methods occurred. The SGLang result is substantial favorable evidence: it reports a low estimated V3 output cost. However, its stated workload limits and the absence of a matched cost account for both original full models prevent a conclusive assessment of the complete claim.
Limitations
- The conclusion should not be read as evidence that DeepSeek's prices required subsidies.
- Requiring corporate profitability would impose a stronger claim than Zitron made.
- The condition is met under the inference-methods reading; a training-reproduction reading remains uncertain.
Evidence (2 items)
Supports the claim: Together AI,
Our endpoint
Why it matters here: The announcement establishes external hosting of full R1. Its cluster prices depend on a specified input/output mix and contract commitments, so they are not directly equivalent to DeepSeek's separate token prices.
- Publisher
- Together AI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: LMSYS Org / SGLang Team,
replicated DeepSeek’s inference
Why it matters here: This establishes recreation of relevant inference methods and external V3 deployment. The authors estimate $0.20 per million output tokens, but identify latency and sequence-length limitations; this favors the claim without settling both full models' commercial economics.
- Publisher
- LMSYS Org / SGLang Team
- Kind of source
- Independent measurement
- Measures
- SGLang Team / Atlas Cloud deployment of DeepSeek-V3, Estimated inference cost under the reported benchmark workload, USD dollars per million output tokens, estimate, 2025-05-05 to 2025-05-05
- Describes events of
- Public from
What would settle it
Reproducible, workload-matched cost accounts for full V3 and R1 that include the relevant serving expenses and compare them with the quoted token prices.
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, Deep Impact
Once these models are hosted elsewhere, and once DeepSeek's methods (which I'll get to shortly) are recreated (which won't take long), I believe we'll see that these prices are indicative of how cheap these models are to run.
- Where
- paragraph 35
- Published
- 2025-01-29
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-deep-impact-44a68b-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.