
12 AI Models That Are No Longer Worth Your Money (Even If You Can Afford Them)
Title history
- 12 AI Models That Are No Longer Worth Your Money (Even If You Can Afford Them) — observed on 20261117-0330 UTC (YouTube)
Keywords
Summary
152 words
Critical Evaluation
Value of the Information & Strength of the Argument
The video provides actionable insights with specific pricing data and benchmarks, making it valuable for developers and businesses. The argumentation is coherent, using examples and comparisons to show how models like GPT-6 Astra are overpriced for routine tasks. It also acknowledges counterpoints, such as the tokenizer issue with Claude Sonnet 5.5 and the contested nature of benchmarks, which strengthens its credibility. However, it relies on a single source for some claims, like the TechCrunch report on GPT-6 Astra.
Scientific Rigor, Source Quality, Title Accuracy
The video cites multiple sources, including Epoch AI, Artificial Analysis, and tech media, but does not provide direct links. The data appears consistent with public knowledge, though some pricing and dates are from the future (2026) which cannot be verified. The title is accurate as it lists exactly 12 models. The content is well-structured and aligns with the title.
152 words
Title / Content Match
The title accurately reflects the content: the video lists exactly 12 AI models and argues they are overpriced with better alternatives.
Quality & Reliability
7/10
The video cites multiple independent sources like Epoch AI and Artificial Analysis, provides specific pricing and benchmark data, and acknowledges contested benchmarks. However, it is a YouTube video with no peer review, and some claims rely on single reports.
Key Moments
Markers derived by PSI from the transcript: the creator did not define chapters.
- Introduction with o1-pro example and the rule about price.
- Number 12: GPT-4.1 with its replacement GPT-6 Luna.
- Number 11: GPT-4o, user backlash and retirement.
- Number 10: Gemini 2.5 Pro, Google's rerouting plan.
- Number 9: Claude Sonnet 4.5, successor Sonnet 5.5.
- Number 7: Deepseek V4 Pro, developer pushback.
- Number 5: o3 Pro, shutdown and replacement.
- Number 1: GPT-6 Astra, why it's number one.
- Five-step audit to reduce costs.
Cited Sources
- Epoch AI — Cited for the finding that the cost of reaching a fixed performance level falls 9 to 900 times a year.
- Artificial Analysis — Cited for intelligence indices and price per task comparisons.
- TechCrunch — Reported on OpenAI scrapping GPT-6.1 Astra.
- Wired — Reported on GPT-4o fans in China.
- Apollo Research — Cited for testing Astra's awareness of being evaluated.
Contribution & Novelties
The video adds value by compiling a current list of overpriced AI models with specific alternatives, helping developers avoid unnecessary costs. It also highlights the psychological and habitual factors that keep users on outdated models.
Pour aller plus loin :
- Large language model — Background on LLMs.
- Artificial Analysis — Independent benchmarking and pricing comparisons.
- Epoch AI — Research on AI trends and cost-performance.
64 words
Radar Profile
The radar indicates high information quantity and quality, with moderate technical depth and reliability. The video is comprehensive and provides actionable advice, though it requires some familiarity with AI pricing and benchmarks.