← The AI Hype Audit — all 645 verdicts
PARTLY
Partly true: the four models are real and cost and speed do vary a lot between them, but '$6 versus $93' for one steam-locomotive prompt comes with no method, no prompt and no way to check the bill
The claim'Same prompt. 4 AI models. 4 very different results. GPT-6 Astra spent $93 and 23 minutes, while Fable 5.2 created a cinematic steam locomotive for just $6 in 2 minutes. Fable 5.2 $6 / 2 min, Opus 5.5 $24 / 7 min, Kimi K3.1 $58 / 16 min, GPT-6 Astra $93 / 23 min. The cheapest workflow came out on top.' Comment PROMPT to get the prompts.
Every model named here exists. GPT-6 Astra, Claude Opus 5.5 and Claude Fable 5.2 were confirmed in earlier checks, and Moonshot released Kimi K3.1 Pro on Oct 23, 2026. The broad lesson, that the priciest model is not always the best pick for a job, is also true and worth knowing. The numbers are the problem. The reel is ten silent seconds of four locomotive scenes with a dollar and minute counter ticking under each. It never shows the prompt, the tool the models ran in, the settings, how the cost was measured or how many tries each got. One detail suggests the totals depend on the setup rather than the models: Kimi K3.1 Pro is the cheapest of the four per token on its public price list (about $2 in, $10 out per million), yet it is shown costing $58. That can happen if a model takes far more steps, but nothing on screen explains it. A single run is also a coin flip; the same prompt can cost very different amounts twice in a row. 'Comment PROMPT' is the lead hook. We could not reproduce or verify any of the four bills.
What holds up
- Contact sheet viewed: four labeled locomotive renders with running cost and time counters, final values matching the caption; no audio.
- Kimi K3.1 Pro confirmed: released Oct 23, 2026, listed at about $2 input and $10 output per million tokens.
- GPT-6 Astra, Opus 5.5 and Fable 5.2 existence carried over from last night's verified entries (x-ref #626, #597).
- Looked for the prompt, harness or billing evidence in the caption and frames: none shown.
What doesn’t
- No prompt, tool, settings or billing screenshot shown.
- One run per model; results and cost vary run to run.
- The cheapest-per-token model is shown as the second most expensive, with no explanation.
- 'Comment PROMPT' is a lead-capture hook.
The catch
Real models and a true general point, wrapped around four dollar figures nobody can check. Treat the ranking as entertainment.
How to actually do it
- Before a big job, run the same small test task on two models and read the actual usage on your own account.
- Set a spending cap or a step limit so a long agent run cannot surprise you.
- Use the cheaper, faster model by default and move up only when it fails the task.
The takeaway survives even if the numbers do not: measure cost on your own work. Nothing in the reel lets a viewer confirm $6 or $93.
- Confidence
- Low
- Posted by
- an AI comparison page (about 11,600 plays); reel sent to Buddy Sun Nov 15, 2026, 7:47 AM
We test hype for free. We build the real thing for a living.
Thirty minutes, no pitch — and you'll leave with something useful either way.
Book a call with Todd or start with the free Business Checkup →The Verdict Weekly
Three verdicts every Friday. Free forever, unsubscribe anytime, no spam — that would be ironic.
© Schreier Group · schreiergroup.com · See a wild AI claim? Drop it here and we'll test it.