GPT-5.4/5.3 family incl Mini/Nano/Pro/Spud; DALL-E 3 deprecation May12/GPT Image 1.5; Image-2 A/B/leaked Arena benchmarks/CarPlay/app integrations; 400k ctx/2x coding/usage +8.9%/quality degradation
Key Questions
What are the key features of the GPT-5.4 family models?
The GPT-5.4 family includes Mini, Nano, Pro, and Spud variants, with a multimodal rollout on March 28. They offer 400k context window, 2x improvements in coding and math, topping Claude benchmarks, and features like $0.75 API and free Thinking mode. Spud is fast-tracked with 20M context for reasoning and drift fixes.
Why has GPT-5.4 usage increased by 8.9%?
Usage spiked post-OpenClaw ban, reaching 45% market share and 900M weekly active users. Developers debate it against GPT-5.3 Codex, with strong performance in coding despite some quality complaints. This follows 1.5M cancellations amid Stanford-reported degradation.
What is the status of DALL-E 3 and its replacement?
DALL-E 3 and DALL-E 2 will be deprecated on May 12, 2026, migrating to GPT Image 1.5 with Edit and outpainting capabilities. GPT Image 1.5 is positioned as the upgrade for image generation. Leaked Image-2 benchmarks show A/B testing in ChatGPT.
What new integrations are available in ChatGPT?
ChatGPT now integrates with Tubi for TV/movie recommendations and direct watching, the first streamer app. Other integrations include CarPlay in iOS 26.5 for voice mode while driving, and apps like DoorDash, Spotify, Uber. These signal a push toward agentic, mobile, and SuperApp features.
What complaints exist about ChatGPT quality degradation?
Users report lazy outputs, drift, and worsening quality, confirmed real by analyses like Atom Writer Blog and Stanford studies. This coincides with 1.5M cancellations despite usage growth. Developer forums debate GPT-5.4 vs. 5.3 Codex amid these issues.
How does GPT-5.4 perform in coding and math?
GPT-5.4 Pro excels in research-level math and doubles coding performance over predecessors, topping Claude in benchmarks. It supports 400k context and is debated in dev forums against GPT-5.3 Codex. Pricing is $200 for Pro with strong API options.
What is GPT Image 2 and its status?
GPT Image 2 is in A/B testing and leaked Arena benchmarks on ChatGPT, following GPT Image 1.5's December 2025 launch. Predictions suggest advanced capabilities based on OpenAI's trajectory. Users may encounter it in tests.
What are the new UI and app features in ChatGPT?
ChatGPT features CarPlay voice mode in iOS 26.5 beta, Tubi app for streaming recs, and broader app integrations. It has 900M WAU with NSFW delay and Instant Checkout trials. These enhance mobile and entertainment use.
GPT-5.4 Mini/Nano/Pro (Mar28 multimodal rollout) usage +8.9% post-OpenClaw ban (400k ctx/2x coding/math tops Claude/$0.75 API/free Thinking/OpenClaw strong/dev forum debates vs 5.3 Codex) but degradation complaints (lazy/drift/1.5M cancellations/45% market share/Stanford); Spud fast-track (20M ctx/reasoning/fix drift); DALL-E 3/2 deprecation May12 to GPT Image 1.5 Edit/outpainting; leaked Image-2 Arena; ChatGPT UI/CarPlay iOS26.5 voice mode/Tubi app (first streamer for TV/movie recs/direct watch)/app integrations/900M WAU/NSFW delay. Benchmarks/integrations signal agentic/image/mobile/SuperApp amid churn.