GPT-5.5 is a total freak
Credibility score: 68/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.
BSmeter analyzed "GPT-5.5 is a total freak" and rated it 68/100 for credibility (a BS score of 32/100 — mostly credible), on 2026-04-27. Its weakest claim — "OpenAI just released GPT-5.5, the best AI model" — scored 0/100 and was flagged as bs. 30 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, the best AI model — BS (0/100)
GPT-5.5? Bro pulled that from fanfic central 💀🔥
Sources: OpenAI releases "Spud" GPT-5.5 model - Axios, GPT-5.5 System Card - OpenAI
HubSpot sponsor shoutout — Sponsored (50/100)
HubSpot plug incoming — fast-forward material 📺💀
GPT-5.5 available on ChatGPT and Codeex app — Verified (95/100)
Dropped GPT-5.5 like it's old news — and damn if it isn't actually real now 😤✅
Codeex lets agents build entire projects in folders — Solid (85/100)
Codeex doing full folder projects? Checks out, but 'way more effective' is pure hype 💀✅
GPT-5.5 can build interactive 3D Earth twin — Personal Story (70/100)
Claims it's 'really hard' but GPT-5.5 just cranks it out — demo sells itself 🔥😤
AI coded interactive 3D globe/map with nightlights, city pans, street view, 3D buildings in few prompts — Personal Story (70/100)
Demo looks slick — standard web tech like CesiumJS/Mapbox, but AI spitting it out in prompts? Impressive flex 😤✅
AI renders 3D scene with accurate material properties in 3 prompts — Solid (80/100)
Sliders doing exactly what they should — reflections, translucency, all spot on. Hate to say it, but this demo slaps ✅😤
ChatGPT has a 'pro 5.5' model selectable in dropdown — Verified (95/100)
GPT-5.5 Pro just dropped April 24 — they're actually using the real deal 😤✅
GPT-5.5 can identify cancer lesions in chest CT images — OK (65/100)
Impressive on obvious lesions but missed the subtle one — not replacing your doc yet 🙄📈
HubSpot's 50+ AI app ideas database making millions, free access — Sponsored (50/100)
Smooth pivot from cancer detection to 'make millions' pitch — classic 💀🤑
HubSpot created free database of succeeding AI apps — Sponsored (50/100)
Classic mid-video sponsor plug — HubSpot database sounds handy but it's an ad 💀📈
Each of six images shows different brain tumor type — Solid (80/100)
Tumor types like meningioma, schwannoma are real — demo setup checks out ✅🧠
State-of-the-art AI can't identify brain tumors from CT — Dubious (45/100)
One failed demo and AI's suddenly useless for tumors? AI crushes this in reality 😤🔥
Other top AI models couldn't code this smooth liquid splash demo — Personal Story (70/100)
Personal test — can't verify but tech exists now so plausible. Still flexing hard 😤✅
No other top AI models generate 3D scenes this detailed — Dubious (45/100)
Bold flex on 'other models can't' — Meshy and Tripo are out here doing detailed 3D from images too 💀
GPT-5.5 crushes other top models on 3D render prompt — Opinion (50/100)
Personal 'not even close' take — cool demo, but we need side-by-side proof 😤
GPT-5.5 made Grammy award-winning quality song — BS (10/100)
'Grammy award-winning' from a prompt? Pump those brakes into next week 💀🔥
LLM composed multi-track music with piano, synth, etc. — Personal Story (70/100)
Demo sounds basic but the multi-track control is legit impressive for a text model 💀✅
Past SOTA models failed at 3D shooter games, only did simple 2D. — Personal Story (65/100)
Personal test history tracks with old LLMs — 3D was their kryptonite back then 😤
GPT-5.5 created fully functional 3D shooter game in two prompts — Personal Story (70/100)
Demo looks legit, AI game gen is real now — but 'two prompts' to AAA level? Chill with the hype 😤✅
No AI has passed frog test; GPT-5.5 failed hidden frog detection — Personal Story (65/100)
'No AI ever passed'? Bold flex on a custom test we can't verify — but COD is still hard 💀🔥
GPT-5.5 optimized for agentic workflows and automation — Verified (90/100)
Damn, GPT-5.5 dropping April 23 and crushing agentic stuff? They weren't lying. Hate when it's legit 😤✅
Codex/GPT-5.5 automates lead scraping and custom site building in minutes — Personal Story (70/100)
Demo shows it spitting out full landing pages with real emails in 3 mins? Impressive if not staged 💀✅
GPT-5.5 does deep Alzheimer's research with detailed analysis and visuals — Personal Story (75/100)
7-min think time for Phase 3 breakdowns with tables? Model's built for this — demo delivers 🔥✅
Tested GPT 5.5, better than Opus 4.7, specs call it smartest model — Personal Story (70/100)
Personal tests on non-existent GPT-5.5? Bold move acting like it's real 💀🔥
GPT-5.5 is smartest model, excels in agentic coding and research — Verified (95/100)
Nailed the capabilities description—straight from OpenAI's announcement. Hate that it's legit 😤✅
GPT-5.5 available to paid Plus/Pro/Business/Enterprise users in ChatGPT/Codex — Solid (85/100)
Access details spot on for paid tiers—free users cry about it 💀✅
GPT-5.5 tops Terminal-Bench, Artificial Analysis; 922K context; beats Claude — Verified (92/100)
Benchmarks and context window check out—leads leaderboards like a boss 📈😤✅
GPT-5.5 xhigh pricier than Opus 4.7 Max, 2x GPT-5.4 xhigh; tops LiveBench. — Solid (80/100)
Pricing math checks out exactly, LiveBench lead too. Hate that it's real 😤✅
GPT-5.5 xhigh tops ARC AGI2 with 85% score. — OK (65/100)
Tops ARC-AGI-2? Sure. But 85%? Pulling that from vibes, not the board 🙄
See the full analysis with sources and timestamps →