Every's vibe check: Gemini 4 Carbon codes better than Argon but judgment lags Fable 5.2
every · x · 2026-11-19
Every published a Vibe Check of Google DeepMind's Gemini 4 Carbon, which launched one day after Anthropic's Fable 5.2:
- Coding: a major step up from Gemini 4 Argon — handles ambitious builds, works for hours with little supervision, and makes better 3D objects.
- Surprise finding: Carbon surfaced a useful pattern in Every's reader survey responses unprompted, grouping readers by how they understand the brand (practical AI use, editorial, products/studio) — closely matching groups the team had identified independently, though the presentation was dense cards-in-cards.
- Judgment gap: testers describe it as working with "a scientist rather than a management consultant" — interesting analysis that needs help communicating; Fable 5.2 has raised the bar on choosing the right work, respecting limits, and knowing when it's done.
A clear win for Google-ecosystem users, but expectations for flagships have shifted.
More from Models
- RadixArk open-sources DaisyRose-27B: 18 days on 144 H200s, Terminal-Bench 5.6% to 12.0% — BanghuaZ · 2026-11-19
- DaisyRose-27B's in-house sandbox: ~80% cheaper than cloud providers, 99.95% success rate — BanghuaZ · 2026-11-19
- GPT-6.2 Astra produces first Lean proof of Smoktunowicz's Gap Theorem in 94 minutes — VraserX · 2026-11-19
- Defending GPT-6.2 Astra: beats Fable 5.2 at chess, produces novel math proofs — VraserX · 2026-11-19
- Is OpenAI's 6.2 Astra a math-specialized 'jagged' model? Reddit theory explained — Neurogence · 2026-11-19
- Map of tabular foundation models shows only two are fully open — GaelVaroquaux · 2026-11-19