Models
0 articles and 2 community signals on Models for people running AI in production, written and curated by Wren with every source linked.
What practitioners are saying
All signals →HN on Gemini 4 Hydrogen: praise for the cost-per-task charts, complaints about names and access routes, and one report of agents overreaching
In the thread on Google's launch post, one commenter says the charts comparing time and cost per task with rival models are the information they look for. Several say the names Hydrogen, Argon and Carbon do not signal size and that there are too many ways to call the model. One self-described Gemini fan says early runs did more than asked or did unrelated work. These are individual accounts.
Why it matters For the leaderboard's practitioner view: early comment is about packaging and how agents fail on long jobs, not benchmark scores, so a trial should use long multi-step tasks and not single prompts.
Discussion on Hacker News →Claude Haiku 5.5: what practitioners say after a day with it
Commenters report near-perfect accuracy on structured classification in production and call it noticeably smarter than GPT-6 Luna, while several argue the 100K-token price cutoff is too low for long agent runs.
Why it matters The cheapest tier is only cheap under 100K tokens; check your prompt sizes before you route to it.
Discussion on Hacker News →