We Tested $200 GPT-5.5 Pro on PhD Level Math
Credibility score: 60/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.
BSmeter analyzed "We Tested $200 GPT-5.5 Pro on PhD Level Math" and rated it 60/100 for credibility (a BS score of 40/100 — mostly credible), on 2026-04-27. Its weakest claim — "OpenAI just released GPT 5.5" — scored 10/100 and was flagged as bs. 22 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Claims analyzed
OpenAI just released GPT 5.5 — BS (10/100)
GPT-5.5 Pro? Straight fanfic. OpenAI's at GPT-4o as of 2026 💀🔥
Sources: GPT-5.5 Complete Guide: Thinking, Pro & 1M Context - Digital Applied
Tested unsolved PhD math on Gemini Ultra, Claude, $200 ChatGPT, now GPT-5.5 Pro — Personal Story (50/100)
Personal testing story on ghost AI — kinda sus but can't disprove their laptop 💀
GPT-5.5 Pro better at solutions, tools, agentic tasks, context, efficiency — Solid (80/100)
Nailed the official claims on agentic boosts and token efficiency. Can't roast facts 😤✅
GPT-5.5 more efficient per API notes; Frontier Math 50%→52.4% T1-3, +1.6% T4 — Dubious (45/100)
API efficiency spot-on but math scores off by a hair—52.4% vs actual 51.7%. Close but sloppy 💀📊
GPT-5.5 Pro solved PhD math in 16 min vs 60+ min for prior model — Personal Story (70/100)
Personal test checks out with known efficiency gains. Hate that it's legit 😤✅
GPT-5.5 Pro suggestions identical to 5.4 Pro on this problem — Personal Story (75/100)
Same outputs, faster delivery. Efficiency win, not genius leap 💀✅
OpenAI hype cycle with frequent .1 version bumps like 5.2 to 5.5 — Opinion (50/100)
Calling out the .1 treadmill — spot on marketing shade 🔥
Testing GPT-5.5 Pro on PhD-level math project — Personal Story (70/100)
Fair test setup for a real PhD math task — sparse non-Hermitian matrices ain't undergrad stuff. Let's see if it delivers 😤✅
GPT-5.5 API calls more expensive than GPT-5.4 — Solid (80/100)
Price check: GPT-5.5 output $30/M vs GPT-5.4's $15/M. Yup, pricier ✅😤
GPT-5.5 performs same as 5.4, just slightly faster — Opinion (50/100)
Personal testing says 'same performance, bit faster' — benchmarks disagree but it's his take 🙄
Comparing GPT-5.4, GPT-5.5, GPT-5.5 Pro on same math task — Solid (80/100)
Solid A/B test plan on real models — time + quality metrics. Actually scientific for a YouTube vid 😤✅
GPT-5.5 Pro uses half tokens for same answer as 5.4 Pro — Personal Story (65/100)
Half tokens for same output? Matches efficiency claims from his tests — intriguing if true 🧐✅
GPT-5.5 Codex took 6:03 vs GPT-5.4's 6:55 on task — Solid (80/100)
Specific timings on a real task — can't fake that without showing the output 💀✅
Testing GPT-5.5 vs 5.4 on agentic coding with VS Code + Codeex — Just Vibes (50/100)
Setting up Codeex in VS Code for coding showdown — this should be good 🔥
GPT-5.5 and 5.4 outputs very similar in quality — Opinion (50/100)
Fair take from someone who uses it daily — outputs matching ain't hype, it's reality 😤
Codex useful for some problems but incapable here — Personal Story (70/100)
PhD-level LaTeX merging exposes AI limits — honest flex from a real user ✅
GPT 5.4 and 5.5 outputs looked rushed and disappointing — Personal Story (70/100)
Fair take from their test — personal results trump benchmarks 😤✅
GPT-5.5 15% more efficient, 1 min less CoT than GPT-5.4 — Solid (80/100)
Their math checks out on the timing gain — aligns with known speedups 📈✅
GPT-5.5 not meaningfully better than 5.4 beyond efficiency — Opinion (50/100)
Bold to dismiss broad improvements based on one test — benchmarks beg to differ 🙄
GPT-5.5 halves the cost — Dubious (45/100)
Halving? API pricing doubled per some sources — check the receipts 💀
GPT-5.5 Pro doesn't bring huge consumer benefit. — Opinion (50/100)
Fair take, but 'Sam Alman's porn'? Auto-caption massacre 💀😂
New DeepSeek model seems more impressive. — Opinion (50/100)
DeepSeek V4 just dropped — personal hype check, but yeah it's stacking up 🔥
See the full analysis with sources and timestamps →