GPT 5.4

1. What are the best OpenAI models for AI agent based on your experiences? (www.reddit.com)

   +23 2w openai

   Hi everyone, I'm torn between using the following models for a financial AI client. It consists of a router client and two sub-clients.

2. Split my agent into a cheap router model and a premium synthesis model, bill dropped about 75% (www.reddit.com)

   +11 3w gpt-5

   I've been building an internal enrichment agent for our team (5 people, B2B sales context) that takes a list of company names and enriches them with public info before our outreach folks touch them. Around 8 tools wired in.

3. Switching from Copilot $39 to Cursor $60 — does it actually make sense long term (www.reddit.com)

   +21 3w copilotcursor

   Hello fellas, I'm here looking for real-life developer experiences using these tools (if you have no programming knowledge and you vibecode 100% of your work, I don't mean to be rude but this post isn't for you). I’m currently on the $39 C…

4. How to integrate AI coding agents to my software (www.reddit.com)

   +11 3w gpt-5copilotgemini+2

   I'm building an locally run application that integrates with coding assistants. So far I've worked with Codex and Copilot.

5. Dynamically allocating compute budget to hard set of problems and evolving the sections with Qwen-35B-A3B gets you near GPT-5.4-xHigh on HLE (www.reddit.com)

   +51 3w gpt-5qwen

   could not extract summary

6. Follow-up to my TranslateGemma-12b benchmark post: human reviewers flagged 71% of the segments automated metrics rated clean (www.reddit.com)

   +2 4w gpt-5deepseeksonnet+1

   A couple of weeks ago I shared the results of a benchmark here showing TranslateGemma-12b beating frontier general models (Claude Sonnet, GPT-5.4, DeepSeek, Gemini Flash Lite) on subtitle translation across 6 languages. The result was stro…

← all threads