Sonnet 5 BACKLASH — Anthropic's New Model Has a Problem

Anthropic just made Sonnet 5 the default for every Claude user — and the reviews are brutal. Here's the honest breakdown: what changed, what broke, and who should actually use it.No fluff. No hype.Sonnet 5 launched with a wave of backlash. Major AI channels called it "horrible" and "the worst model Anthropic ever made." But beneath the noise, there's a more interesting story: Sonnet 5 is Anthropic's first agentic Sonnet — it can plan, use browsers and terminals, and run autonomously at a level Sonnet 4.6 never could. It's just... also more expensive, less efficient in certain tasks, and nowhere near the top 10 on leaderboards. In this video, I go through exactly what changed — the new tokenizer (1-1.35x more tokens per input), the introductory vs post-intro pricing ($2M/$10M vs $3M/$15M per token), the benchmarks vs Opus 4.8, the hands-on invoice builder test, and my honest take on who should switch and who should stay on Sonnet 4.6. ⏱ 00:00 — Sonnet 5 Backlash ⏱ 00:54 — Pricing and Tokenizer Changes ⏱ 02:36 — Benchmarks and Default Switch ⏱ 04:02 — Agentic Browser and Computer Use ⏱ 05:02 — Mixed Reviews and Token Costs ⏱ 06:59 — Leaderboards and Cost per Task ⏱ 08:18 — Hands On Invoice Builder Test ⏱ 10:40 — Who Should Use Which Model ⏱ 12:34 — Final Take and Next Steps My honest take: If you're a simple business user doing knowledge work, it's fine — you're probably already using it. If you're a heavy coder or builder, stick with Sonnet 4.6 or Opus 4.8. The browser use is genuinely better, but the token efficiency tradeoff is real. Cost per task: 6,015 vs Opus 4.8 at 3,753. 📎 Resources: 🔗 Build n Bloom Skool Community — https://skool.com/@build-n-bloom-5406 💬 Drop a comment: Are you sticking with Sonnet 5 as your daily driver, or going back to 4.6? 👇 #ClaudeSonnet5 #Anthropic #AINews #ClaudeAI #Sonnet5Review #AITools #AIforBusiness #SMEtips