Credibility score: 51/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "Gemma 4 12B Is INSANE – Is THIS the BEST Local Coding Model Yet?" and rated it 51/100 for credibility (a BS score of 49/100 — mixed credibility), on 2026-06-04. Its weakest claim — "Claims the app has a section called 'Asian skills' with 10 built-in skills" — scored 25/100 and was flagged as sketchy. 14 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Of 14 claims analyzed: 2 scored under 40, 10 between 40 and 69, and 2 at 70 or above.
At 0:30
No trace in current coverage — might be early or niche.
Why this score: - Web results focus on model specs and local running, not a dedicated macOS desktop app - Speaker says they're testing it, so possible but unconfirmed publicly
Original quote: “Now something else interesting is they have also released a new Mac OS desktop experience”
At 2:30
✅
Why this score: - Matches technical descriptions of Gemma 4 12B as unified encoder-free multimodal model. - Community sources confirm 12B runs locally without additional encoders for images/audio.
Original quote: “So this model the 12B having basically nothing for those ties back into what is being referred to right here.”
At 8:30
Calling it 'insane' twice in 10 seconds — hype dial at 11 already 🔥
Why this score: Subjective hype language. - Title itself uses 'INSANE' so this is just repeating the framing - No benchmarks shown yet at this timestamp - Community comments show people are waiting for more models, not declaring this one the winner
Original quote: “This model's getting insane in the title. No if ends or buts about it. This model is getting insane in the title.”
At 10:30
✅
Why this score: - Video title and description confirm Apache 2.0 license - No contradictory info in search results
Original quote: “should have also mentioned in the beginning that this is Apache 2.0. So if there are any comments that say like you didn't mention Apache 2.0, just know that they didn't watch the video thoroughly.”
At 10:48
Subjective visual judgment — no objective metric given
Why this score: - Purely qualitative assessment of the output - No benchmark scores or error rates provided
Original quote: “For a model of this size, I am okay with this. I know that may sound weird, but it's doing it in the correct shape.”
At 11:10
Viewer experience, not measurable claim
Why this score: - Personal observation of interaction quality - No quantitative frame-rate or latency data
Original quote: “It's an ultra sparse print. The movement though of the camera, like if I click and rotate like here. Okay, I'm clicking, spin, let go. It's got nice fluidity to the movement, which I noticed in the GTA sim as well. This model's smooth.”
At 12:30
Subjective judgment call — no objective metric given.
Why this score: • Speaker is rating visual fidelity by eye. - No side-by-side comparison or quantitative score provided. - Claim stays within personal evaluation bounds.
Original quote: “the resemblance does exist in terms of the color palettes it's chosen to use, the relative orientation of the folks in the photo and the expressions. You know, it showed good ability to capture uh color palette and maybe composition of said image if not one to one.”
At 16:30
Either a typo or auto-correct disaster — 'Asian skills' isn't a thing 💀
Why this score: No evidence of any 'Asian skills' category. - Likely a transcription or speech error for 'Asian' instead of something like 'advanced' or app-specific branding - No documentation or community mentions support this label
Original quote: “It does have this specific Asian skills section where if we click on skills right here, we can see there are 10 specific built-in skills.”
At 18:30
Pure opinion, no fact to check
Why this score: Subjective judgment about UI quality. No measurable claim to verify.
Original quote: “for a 12 billion parameter model, this is not a bad front end at all”
At 20:30
Rewriting files when edits fail is basic behavior, not freakish
Why this score: Overstated reaction. - Models routinely fall back to full rewrites on compile failure - No comparison to other 12B models shown - "Dumbfounded" reaction doesn't match the actual behavior
Original quote: “This model is insane. Insane. Look at this. Please don't I am I am dumbfounded. Okay, I am dumbfounded that this that this worked because it was making errors. it couldn't figure out how to edit like the single lines or it was messing up with that and then it just decided I'm just going to rewrite…”
At 22:30
Classic hype line — everyone says their model 'democratizes' AI
Why this score: Marketing speak. - 12B models still need decent GPUs; not truly 'mega computer' free - Community posts confirm 12B runs on 16GB laptops but performance varies - Phrase is subjective, not measurable
Original quote: “This will democratize intelligence to those who do not have like mega computers at home, which is awesome.”
At 24:30
Cost comparison asserted with no actual pricing or token counts shown ⚖️
Why this score: Missing numbers. - No GPT-4o or API pricing referenced - No local inference electricity/hardware cost included - Single anecdote presented as general rule
Original quote: “So say I had chat GPT create this, which would have been expensive in the scheme of things versus just doing this locally on my system, but we had Gemma 412b do it and then we had chat GPT clean up like a couple of bug fixes in a smaller pass.”
At 26:30
Just recounting his workflow — no bold claim to check
Why this score: Speaker describing his own test session. Pure anecdote, zero external fact to verify.
Original quote: “in the meantime I had Open Code take a stab at the subway scene that we had received and for 10 minutes it was working on turning that into a very cool FPS. I don't know if it's going to work or not.”
At 30:30
Bold crown for a single-test model.
Why this score: Subjective ranking. - No public benchmark leaderboard puts Gemma 4 12B #1 across coding evals. - "Best" depends on hardware, quantization, and task; community tests show it competitive but not undisputed.
Original quote: “I would believe when factoring in being able to be run locally with raw capability, I would say this is perhaps the best local coding model that currently exists”