Last Week in AI #335 - Opus 4.6, Codex 5.3, Gemini 3 Deep Think, GLM 5, Seedance 2.0
Last Week in AI Last Week in AI ● Covered by 2 sources
Last Week in AI's newsletter #335 rounds up a wild stretch of AI releases: Opus 4.6, Codex 5.3, Gemini 3 Deep Think, GLM 5, and Seedance 2.0. The editor also promises steadier weekend publishing and a new commentary feature going forward.
Based on reporting by Last Week in AI, Last Week in AI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Five major model updates dropped in roughly the same window, and the team behind Last Week in AI decided the only sane response was to cram them all into one supersized edition, #335. Opus 4.6, Codex 5.3, Gemini 3 Deep Think, GLM 5, and Seedance 2.0 all show up in the same newsletter, which tells you something about the pace labs are shipping at right now — nobody is waiting for a quiet news cycle anymore.
The editor opens with an apology, actually, for months of inconsistent release timing on both the newsletter and the podcast. That's a small but telling admission: even a publication whose entire job is tracking AI's breakneck schedule has struggled to keep its own schedule steady. The fix going forward is a simple one — aim for Saturday or Sunday, every time, no more drift.
There's also a structural change worth flagging. Starting with this edition, the Top News section gets an 'Editor's Take,' a spot for actual commentary and added context rather than just a rundown of headlines. Given how crowded this particular week was — five flagship-level releases from what looks like five different labs, spanning coding assistants, deep reasoning tools, and video generation — that kind of framing is probably overdue. Raw lists of model names stop being useful once there are this many of them landing in a single fortnight.
What stands out isn't any single release here so much as the sheer density of them arriving together. Text models, coding models, reasoning-focused variants, and a video generator all updating in the same stretch suggests the major labs are no longer staggering announcements to avoid stepping on each other — they're just pushing whenever they're ready, competitors be damned.
My take — AI-written commentary, not fact-checked reporting
Five headline model launches in one newsletter cycle isn't sustainable coverage, it's a symptom — labs are racing each other so hard that nobody bothers to space out announcements anymore, and the actual capability differences between something like Opus 4.6 and GLM 5 get lost in the noise. I'd rather see fewer, more transparent releases with real benchmarks than this firehose approach that rewards whoever ships first, not whoever ships best.
Read more about this at: Last Week in AI