GPT-5.5

Willison used Fable to do a pre-release review of sqlite-utils 4.0rc2: 37 prompts, 34 commits, five blocker bugs found — the worst, delete_where() leaving a transaction open so later writes silently rolled back. He estimates $149.25 in API cost; a cross-review by GPT-5.5 caught two more.

The reversal: $150 spent by an expert who knows exactly what to ask is not the $150 a novice will spend chasing invented problems — because these models never say "nothing found, good to go."

Across 390,195 logged responses, GPT-5.5 Codex terminates reasoning at exactly 516 tokens far more than any sibling: 82% of exact-516 events are GPT-5.5, though it's only 19% of traffic. The stops cluster at 516, 1034, and 1552, each about 518 apart.

One HN user reproduced it live: 4 of 10 identical prompts short-circuited at 516 tokens, and every one returned the wrong answer.

The suspected cause: a throughput optimization batching reasoning in multiples of 512.

On Cursor's numbers, Composer 2.5 hits 63.2% accuracy at $0.55 a task — the cheapest in the field, and framed as near-frontier. The presentation even flipped the cost axis so cheapest sits on the right, which several readers found conveniently confusing.

The rebuttal came with citations. On the DeepSWE benchmark — 'probably the hardest to game' — GPT-5.5 gets 64, Opus 4.8 gets 56, and Composer 2.5 gets 16. A developer who's used both models heavily called the parity claim 'absolutely farcical': faster, yes; anywhere near as good, no.

Computer use is now native in Gemini 3.5 Flash: the model sees a screenshot, reasons, and acts across browser, mobile and desktop, pitched for software testing and enterprise automation via the API and a Browserbase demo.

Then the thread turns on the marketing. Readers note the headline benchmark chart visually crowns Flash while the numbers show two rivals ahead, and agent builders argue screenshot-driving is the wrong primitive entirely.

Show HN "Are You in the Weights?" fires your name at a dozen-plus models, from Llama 3.2 1B to GPT-5.5, and grades how 'memorized' you are with a strength score and a 'Top N%' rank. The replies read like funhouse mirrors: people with a genuinely unique name still got fake biographies, fake sports careers, fake crimes.

Then a privacy warning surfaced: every name typed in lands on a public 'latest' leaderboard that paginates endlessly. And one quiet question nobody answered: what does being in the weights mean for the right to be forgotten?

Z.ai's GLM-5.2 (744B total, 40B active) is now the top open-weights model on Artificial Analysis at 51, ahead of DeepSeek V4 Pro and Kimi K2.6, and it scored 1524 on one benchmark versus GPT-5.5's 1514.

The catch is verbosity: 43k tokens per task, 37k of it reasoning. One tester clocked 15 minutes and 45k tokens before the first file appeared. Sibylline's chief scientist still recommends it — pair GLM for writing with a cheap GPT sub for review, and you've got near-Opus output at a fraction of the price.