{"slug": "gpt-image", "name": "GPT Image", "category": "ai-models", "category_name": "AI models", "description": "", "recommenders": 10, "commenters": 27, "warners": 6, "score": 4.416, "rising": 1.38, "first": "2025-11-20", "latest": "2026-10-15", "why": "The best model I've found for this, that almost bakes in full ControlNet capability, is oddly gpt-image-1.5.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 19, "neutral": 10, "compare": 6, "warn": 9}, "spark": [0, 2, 1, 0, 0, 3, 3, 2, 3, 2, 3, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "The best model I've found for this, that almost bakes in full ControlNet capability, is oddly gpt-image-1.5.", "kind": "", "icon": false, "family": "gpt", "domain": "chatgpt.com", "github_repo": null, "official_url": "", "pricing": "", "aliases": ["gpt-image-2", "GPT Image 2", "GPT Image 1.5", "gpt-image-1", "GPT-Image-2", "GPT Image", "gpt-image-1.5", "GPT Image 1", "GPT-image-1", "GPT-image-1.5", "GPT-Image", "GPT-image-2"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 793868848, "item_id": 49616803, "url": "https://news.ycombinator.com/item?id=49616803", "by": "vunderba", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "gpt-image-2 got it right every time.", "excerpt": "That\u2019s surprising even gpt-image-1 was able to handle the equestrian astronaut prompt pretty trivially, albeit yellow-tinged as all hell. For the heck of it, I ratcheted up the difficulty of the prompt (two-headed horse, astronaut with the visor up, etc.), and gpt-image-2 got it right every time. https://imgpb.com/HwDVcJo", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793862577, "item_id": 49616411, "url": "https://news.ycombinator.com/item?id=49616411", "by": "jjcm", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "Right now gpt-image-2 is better than my own model, and it services around ~96% of the requests in diffui.", "excerpt": "It's both. I have several models loaded into diffui. Which one each prompt uses is determined based on user preference over time. Whichever image is currently at the top of any image node stack marks a win for the model that generated it, I assign each model an ELO score based on that, and I bias the chance each model is selected based on their ELO score. Right now gpt-image-2 is better than my own model, and it serv", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793853840, "item_id": 49615865, "url": "https://news.ycombinator.com/item?id=49615865", "by": "jjcm", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "I use gpt image 2 very heavily for my current project (ai UI design tool).", "excerpt": "I use gpt image 2 very heavily for my current project (ai UI design tool). The biggest improvement I'm seeing with this is in speed. I've generated around 50k images with gpt-image-2 via api, the the average latency has held at around 104s. It's wild how much of a difference this is - images are coming in at around 35-40s. Very noticable, and makes a difference when you're iterating quickly: https://jjcm.org/gpt-imag", "context": "building an AI UI design tool", "use_case": "AI UI design image generation", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793844817, "item_id": 49615301, "url": "https://news.ycombinator.com/item?id=49615301", "by": "minimaxir", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "it crushed every complex image comprehension benchmark I had", "excerpt": "A big wtf at the LM Arena scores: https://arena.ai/leaderboard/text-to-image gpt-image-2.5-sunburst: 1421 gpt-image-2.5-flare: 1399 gpt-image-2 (medium): 1381 mai-image-2.6: 1331 Even with LM Arena being flawed, this is significant. I was planning to do a writeup on the original gpt-image-2 as it crushed every complex image comprehension benchmark I had...I'm glad I procrastinated since ChatGPT Images 2.5 seems like ", "context": "", "use_case": "image generation", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793443729, "item_id": 49590233, "url": "https://news.ycombinator.com/item?id=49590233", "by": "heaney-555", "date": "2026-09-06", "version": "2", "stance": "recommend", "quote": "the same prompt in GPT-Image-2 has no such issue", "excerpt": "You are conflating the specific technical limitations of Nano Banana 2 with the general nature of AI. This is akin to the \"God of the gaps\" fallacy, wherein your worldview is going to be repeatedly decimated by advancements in models. Nano Banana 2 isn't even in the top 5 image models anymore - it's obsolete: https://arena.ai/leaderboard/text-to-image For example, the same prompt in GPT-Image-2 has no such issue (I e", "context": "", "use_case": "", "switched_from": "", "story": "GPT-6 Astra"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 793868849, "item_id": 49616803, "url": "https://news.ycombinator.com/item?id=49616803", "by": "vunderba", "date": "2026-09-08", "version": "1", "stance": "warn", "quote": "albeit yellow-tinged as all hell", "excerpt": "That\u2019s surprising even gpt-image-1 was able to handle the equestrian astronaut prompt pretty trivially, albeit yellow-tinged as all hell. For the heck of it, I ratcheted up the difficulty of the prompt (two-headed horse, astronaut with the visor up, etc.), and gpt-image-2 got it right every time. https://imgpb.com/HwDVcJo", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793860769, "item_id": 49616298, "url": "https://news.ycombinator.com/item?id=49616298", "by": "jjcm", "date": "2026-09-08", "version": "2", "stance": "warn", "quote": "Overall fixed some of the main issues / gripes I had with gpt-image-2", "excerpt": "Some UI tests with it: Warcraft 3 style agentic dev interface: https://image.non.io/cd9ea5cd-8ed7-44e0-ad3f-480ff0e51875.webp Overall it used the reference images I gave it a bit better than gpt-image-2. I noticed 2 had issues getting the blue button just right. 2.5 nailed it. A \"John Politics\" meme site: https://image.non.io/d2922164-fa96-4d07-a141-2febadb02939.webp Did very well modifying the pose while keeping the", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793845424, "item_id": 49615339, "url": "https://news.ycombinator.com/item?id=49615339", "by": "vunderba", "date": "2026-09-08", "version": "2", "stance": "warn", "quote": "their 4-panel comic strip using gpt-image-2 was an absolute mess too", "excerpt": "The OpenAI PR/doc team seem to have a history of this kind of attention lapse - their 4-panel comic strip using gpt-image-2 was an absolute mess too. https://mordenstar.com/blog/gen-failures", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 769216080, "item_id": 48076005, "url": "https://news.ycombinator.com/item?id=48076005", "by": "vunderba", "date": "2026-05-09", "version": "2", "stance": "warn", "quote": "It\u2019s not a great look when the biggest proponents and supposed experts still manage to make such a royal botch job of things.", "excerpt": "An even bigger issue to me is that people who most frequently tend to avail themselves of AI art also tend to be the least observant. So even when the AI-generated image could have been good or at least serviceable, they somehow still manage to bungle the whole process. I actually wrote a blog post about this, with concrete examples (and my manual fixes) taken from OpenAI when they demoed gpt-image-2 and from Karpath", "context": "", "use_case": "", "switched_from": "", "story": null}]}], "who": [{"by": "jjcm", "context": "building an AI UI design tool", "url": "https://news.ycombinator.com/item?id=49615865", "date": "2026-09-08"}, {"by": "vunderba", "context": "runs a GenAI image model benchmark", "url": "https://news.ycombinator.com/item?id=47938854", "date": "2026-04-28"}], "tags": [], "use_cases": ["image generation", "ai image generation", "image generation and editing", "image generation with alpha", "ai ui design image generation", "image generation for design"], "compared": [{"slug": "nano-banana-pro", "name": "Nano Banana Pro", "co": 10, "switched_from": 0, "switched_to": 0}, {"slug": "nano-banana", "name": "Nano Banana", "co": 8, "switched_from": 0, "switched_to": 0}, {"slug": "codex", "name": "Codex", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "chatgpt", "name": "ChatGPT", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "claude", "name": "Claude", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "qwen-image", "name": "Qwen-Image", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [{"id": 740946064, "item_id": 46309129, "url": "https://news.ycombinator.com/item?id=46309129", "by": "yuni_aigc", "date": "2025-12-17", "version": "1.5", "stance": "recommend", "quote": "I built GPT Image 1.5 (https://gpt-image-15.com), a web interface", "excerpt": "Hey HN! I built GPT Image 1.5 (https://gpt-image-15.com), a web interface for OpenAI's latest image generation model released Dec 16, 2025. **What it does:** - Text-to-image generation with precise instruction following - Targeted image editing (e.g., \"change shirt color\" without touching face/background) - Superior text rendering in images (infographics, UI mockups, marketing materials) - 4x faster than previous mod", "context": "", "use_case": "image generation web UI", "switched_from": "", "story": "Built a web UI for OpenAI's GPT Image 1.5 (Next.js 16 and React 19)"}], "threads": [{"url": "https://news.ycombinator.com/item?id=47852835", "title": "OpenAI Livestream", "mentions": 9}, {"url": "https://news.ycombinator.com/item?id=49614720", "title": "ChatGPT Images 2.5", "mentions": 7}, {"url": "https://news.ycombinator.com/item?id=46291941", "title": "GPT Image 1.5", "mentions": 7}, {"url": "https://news.ycombinator.com/item?id=49183850", "title": "Qwen 3.0 Image Pro", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=47853000", "title": "ChatGPT Images 2.0", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=47209694", "title": "Show HN: Imagedojo.ai \u2013 Blind arena for Google, OpenAI, and xAI image generators", "mentions": 2}], "sentiment_by_month": {"2025-11": {"compare": 1}, "2025-12": {"recommend": 2, "neutral": 2, "warn": 3}, "2026-01": {"recommend": 1}, "2026-02": {"compare": 1, "warn": 1}, "2026-03": {"warn": 1}, "2026-04": {"warn": 2, "neutral": 4, "recommend": 3, "compare": 2}, "2026-05": {"recommend": 3, "neutral": 2}, "2026-06": {"recommend": 2}, "2026-07": {"warn": 1, "recommend": 3, "neutral": 1}, "2026-08": {"compare": 1, "recommend": 2}, "2026-09": {"recommend": 3, "warn": 1}, "2026-10": {"neutral": 1, "compare": 1}}, "versions": [{"v": "2", "people": 17, "rec": 7, "warn": 2, "other": 8, "first": "2026-04-21"}, {"v": "1.5", "people": 7, "rec": 3, "warn": 4, "other": 0, "first": "2025-12-16"}, {"v": "1", "people": 5, "rec": 0, "warn": 1, "other": 4, "first": "2025-12-06"}], "version_unnamed": 3, "family_ref": {"slug": "gpt", "name": "GPT"}, "runs_on": null, "lines": [], "quotes": [{"id": 801715250, "item_id": 50107203, "url": "https://news.ycombinator.com/item?id=50107203", "by": "CuriouslyC", "date": "2026-10-15", "version": null, "stance": "compare", "quote": "the GPT-Image look will become blase from overuse", "excerpt": "As a Krea2 local model enjoyer, your challenge is that on the $20 GPT plan you get a fairly good amount of Codex use and image gen. If I used Krea, it would be because I fine tuned/built a workflow on top of it to do something GPT image couldn't/wouldn't do. Maybe there's a market for the cheapest \"good enough\" bulk image gen model (spam? idk what the applications would be), but personally I'd say you should focus on", "context": "", "use_case": "image generation", "switched_from": "", "story": "Krea 3 \u2013 An image model built for editing at scale"}, {"id": 799712784, "item_id": 49982049, "url": "https://news.ycombinator.com/item?id=49982049", "by": "neom", "date": "2026-10-06", "version": "2", "stance": "neutral", "quote": "Here is the same prompt from GPT Image 2", "excerpt": "\"A macro close-up photograph of an old watchmaker's hands carefully replacing a tiny gear inside a vintage pocket watch. The watch mechanism is partially submerged in a shallow dish of clear water, causing visible refraction and light caustics across the brass gears. A single drop of water is falling from a pair of steel tweezers, captured mid-splash on the water's surface. Reflect the watchmaker's face, slightly dis", "context": "", "use_case": "", "switched_from": "", "story": null}, {"id": 796430353, "item_id": 49776897, "url": "https://news.ycombinator.com/item?id=49776897", "by": "jjcm", "date": "2026-09-20", "version": "2", "stance": "compare", "quote": "comparing the outputs of gpt-image-2 and qwen 2.1", "excerpt": "I run a prompt-to-ui design site that uses image models for the design process[1]. The text rendering especially makes this model deeply interesting to me, despite the license. Here are some tests using my harness comparing the outputs of gpt-image-2 and qwen 2.1: https://html.non.io/qwen-comparison/ The text rendering definitely is much, much better than anything else on the open weights market right now. Small text", "context": "", "use_case": "", "switched_from": "", "story": "Qwen Image 2.1"}, {"id": 793868848, "item_id": 49616803, "url": "https://news.ycombinator.com/item?id=49616803", "by": "vunderba", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "gpt-image-2 got it right every time.", "excerpt": "That\u2019s surprising even gpt-image-1 was able to handle the equestrian astronaut prompt pretty trivially, albeit yellow-tinged as all hell. For the heck of it, I ratcheted up the difficulty of the prompt (two-headed horse, astronaut with the visor up, etc.), and gpt-image-2 got it right every time. https://imgpb.com/HwDVcJo", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793868849, "item_id": 49616803, "url": "https://news.ycombinator.com/item?id=49616803", "by": "vunderba", "date": "2026-09-08", "version": "1", "stance": "warn", "quote": "albeit yellow-tinged as all hell", "excerpt": "That\u2019s surprising even gpt-image-1 was able to handle the equestrian astronaut prompt pretty trivially, albeit yellow-tinged as all hell. For the heck of it, I ratcheted up the difficulty of the prompt (two-headed horse, astronaut with the visor up, etc.), and gpt-image-2 got it right every time. https://imgpb.com/HwDVcJo", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793862577, "item_id": 49616411, "url": "https://news.ycombinator.com/item?id=49616411", "by": "jjcm", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "Right now gpt-image-2 is better than my own model, and it services around ~96% of the requests in diffui.", "excerpt": "It's both. I have several models loaded into diffui. Which one each prompt uses is determined based on user preference over time. Whichever image is currently at the top of any image node stack marks a win for the model that generated it, I assign each model an ELO score based on that, and I bias the chance each model is selected based on their ELO score. Right now gpt-image-2 is better than my own model, and it serv", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793860769, "item_id": 49616298, "url": "https://news.ycombinator.com/item?id=49616298", "by": "jjcm", "date": "2026-09-08", "version": "2", "stance": "warn", "quote": "Overall fixed some of the main issues / gripes I had with gpt-image-2", "excerpt": "Some UI tests with it: Warcraft 3 style agentic dev interface: https://image.non.io/cd9ea5cd-8ed7-44e0-ad3f-480ff0e51875.webp Overall it used the reference images I gave it a bit better than gpt-image-2. I noticed 2 had issues getting the blue button just right. 2.5 nailed it. A \"John Politics\" meme site: https://image.non.io/d2922164-fa96-4d07-a141-2febadb02939.webp Did very well modifying the pose while keeping the", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793853840, "item_id": 49615865, "url": "https://news.ycombinator.com/item?id=49615865", "by": "jjcm", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "I use gpt image 2 very heavily for my current project (ai UI design tool).", "excerpt": "I use gpt image 2 very heavily for my current project (ai UI design tool). The biggest improvement I'm seeing with this is in speed. I've generated around 50k images with gpt-image-2 via api, the the average latency has held at around 104s. It's wild how much of a difference this is - images are coming in at around 35-40s. Very noticable, and makes a difference when you're iterating quickly: https://jjcm.org/gpt-imag", "context": "building an AI UI design tool", "use_case": "AI UI design image generation", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793845424, "item_id": 49615339, "url": "https://news.ycombinator.com/item?id=49615339", "by": "vunderba", "date": "2026-09-08", "version": "2", "stance": "warn", "quote": "their 4-panel comic strip using gpt-image-2 was an absolute mess too", "excerpt": "The OpenAI PR/doc team seem to have a history of this kind of attention lapse - their 4-panel comic strip using gpt-image-2 was an absolute mess too. https://mordenstar.com/blog/gen-failures", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793844817, "item_id": 49615301, "url": "https://news.ycombinator.com/item?id=49615301", "by": "minimaxir", "date": "2026-09-08", "version": "2", "stance": "recommend", "quote": "it crushed every complex image comprehension benchmark I had", "excerpt": "A big wtf at the LM Arena scores: https://arena.ai/leaderboard/text-to-image gpt-image-2.5-sunburst: 1421 gpt-image-2.5-flare: 1399 gpt-image-2 (medium): 1381 mai-image-2.6: 1331 Even with LM Arena being flawed, this is significant. I was planning to do a writeup on the original gpt-image-2 as it crushed every complex image comprehension benchmark I had...I'm glad I procrastinated since ChatGPT Images 2.5 seems like ", "context": "", "use_case": "image generation", "switched_from": "", "story": "ChatGPT Images 2.5"}, {"id": 793443729, "item_id": 49590233, "url": "https://news.ycombinator.com/item?id=49590233", "by": "heaney-555", "date": "2026-09-06", "version": "2", "stance": "recommend", "quote": "the same prompt in GPT-Image-2 has no such issue", "excerpt": "You are conflating the specific technical limitations of Nano Banana 2 with the general nature of AI. This is akin to the \"God of the gaps\" fallacy, wherein your worldview is going to be repeatedly decimated by advancements in models. Nano Banana 2 isn't even in the top 5 image models anymore - it's obsolete: https://arena.ai/leaderboard/text-to-image For example, the same prompt in GPT-Image-2 has no such issue (I e", "context": "", "use_case": "", "switched_from": "", "story": "GPT-6 Astra"}, {"id": 786991441, "item_id": 49186965, "url": "https://news.ycombinator.com/item?id=49186965", "by": "jjcm", "date": "2026-08-05", "version": "2", "stance": "recommend", "quote": "it's nowhere near gpt-image-2 at least for high density UI design", "excerpt": "I was playing with this a bunch when it dropped a week ago. It's a big step up over their prior model, but it's nowhere near gpt-image-2 at least for high density UI design. Photos are a solved domain in my eyes, so the real question is if it can do infographics/web design. Here are some samples, each with the same prompt: Cannabis site gpt-image-2: https://image.non.io/9fbf3396-889a-445e-b51b-ab4468ded269.webp qwen-", "context": "", "use_case": "", "switched_from": "", "story": "Qwen 3.0 Image Pro"}, {"id": 786968560, "item_id": 49185535, "url": "https://news.ycombinator.com/item?id=49185535", "by": "vunderba", "date": "2026-08-05", "version": null, "stance": "recommend", "quote": "It\u2019s easier to correct an issue like this than to deal with an image that simply doesn\u2019t follow your prompt.", "excerpt": "True but it\u2019s at least addressable with some basic tone-mapping changes. It\u2019s easier to correct an issue like this than to deal with an image that simply doesn\u2019t follow your prompt. Here's a quick trend of the \"piss filter\" in the gpt image series: https://imgpb.com/vCZidh", "context": "", "use_case": "image generation", "switched_from": "", "story": "Qwen 3.0 Image Pro"}, {"id": 786959872, "item_id": 49184992, "url": "https://news.ycombinator.com/item?id=49184992", "by": "sinak", "date": "2026-08-05", "version": "2", "stance": "compare", "quote": "Looks like this model is meaningfully less good than gpt-image-2", "excerpt": "Looks like this model is meaningfully less good than gpt-image-2. Arena.ai score is 1263 vs 1380. https://arena.ai/leaderboard/text-to-image", "context": "", "use_case": "", "switched_from": "", "story": "Qwen 3.0 Image Pro"}, {"id": 785881040, "item_id": 49117565, "url": "https://news.ycombinator.com/item?id=49117565", "by": "jjcm", "date": "2026-07-30", "version": "2", "stance": "recommend", "quote": "Gpt-image-2 is a VERY capable designer", "excerpt": "> There are other aesthetics my brain associates with AI, like beige/cream colors, orange accents, and serif typefaces Something to consider regarding the narrow space AI-created designs align on: LLMs are trained to write consistent code. This makes sense for something like a billing or a backend function - you want that code to be consistent. The problem though is that LLMs write code to represent designs as well, ", "context": "Left Figma to build a design tool", "use_case": "Image generation for design", "switched_from": "", "story": "The AI Aesthetic"}, {"id": 784611043, "item_id": 49038190, "url": "https://news.ycombinator.com/item?id=49038190", "by": "vunderba", "date": "2026-07-24", "version": "2", "stance": "compare", "quote": "gpt-image-2 scored 12 out of 15", "excerpt": "I run a fairly high-traffic site for generative image models focusing on complex prompt adherence. Flux.2 doesn\u2019t score anywhere near SOTA proprietary models. Klein 9b is decent for image-to-image, but when used for pure generative purposes brings back SDXL levels of body horror (have some Gattica pianists). If you can get past the annoying JSON structuring, Ideogram 4 is probably the best option in the open\u2011weights ", "context": "", "use_case": "", "switched_from": "", "story": "Flux 3"}, {"id": 783840128, "item_id": 48990008, "url": "https://news.ycombinator.com/item?id=48990008", "by": "postalcoder", "date": "2026-07-21", "version": "1", "stance": "neutral", "quote": "They must have trained on GPT Image 1 outputs.", "excerpt": "They must have trained on GPT Image 1 outputs. The yellow tint is unmistakable. https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen-Image/image3/whale.jpg https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen-Image/image3/update5.png https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen-Image/image3/mlp.png https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen-Image/image3/image%20-%202026-07-21T003745.645.png", "context": "", "use_case": "", "switched_from": "", "story": "Qwen-Image-3.0: Rich Content, Authentic Details, Deep Knowledge"}, {"id": 783693361, "item_id": 48980835, "url": "https://news.ycombinator.com/item?id=48980835", "by": "vunderba", "date": "2026-07-20", "version": "2", "stance": "recommend", "quote": "You could also use a multimodal model such as gpt-image-2", "excerpt": "Not OP, but it looks to be AI generated which was then heavily downscaled in size, since if you zoom in you can still see the usu. telltale suspects such as weird purple/pink/blue color artifacts, etc. Here's some quick similar examples I generated using Ideogram v4 running locally on my machine. You could also use a multimodal model such as gpt-image-2, give it a specific example jpg (e.g. tiger swallowtail butterfl", "context": "", "use_case": "pixel art generation", "switched_from": "", "story": "Show HN: Stacks \u2013 an hourly card solitaire about building four gardens"}, {"id": 783233120, "item_id": 48952070, "url": "https://news.ycombinator.com/item?id=48952070", "by": "simonw", "date": "2026-07-17", "version": "2", "stance": "recommend", "quote": "Here's one I got from OpenAI's gpt-image-2 just the other day", "excerpt": "The dedicated text-to-image models all produce good illustrations of pelicans riding bicycles. Here's one I got from OpenAI's gpt-image-2 just the other day: https://simonwillison.net/2026/Jul/14/pedalican/", "context": "", "use_case": "Image generation", "switched_from": "", "story": "Kimi K3, and what we can still learn from the pelican benchmark"}, {"id": 782376192, "item_id": 48898512, "url": "https://news.ycombinator.com/item?id=48898512", "by": "throwthrowuknow", "date": "2026-07-13", "version": "2", "stance": "warn", "quote": "", "excerpt": "It\u2019s a tough problem but Retro Diffusion is making some incredible progress https://retrodiffusion.ai/", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Super Dario"}, {"id": 778112144, "item_id": 48632009, "url": "https://news.ycombinator.com/item?id=48632009", "by": "vunderba", "date": "2026-06-22", "version": "2", "stance": "recommend", "quote": "Proprietary? Either gpt-image-2 or NB2.", "excerpt": "Proprietary? Either gpt-image-2 or NB2. I have an example of interior decorating inpainting where I replaced a large floor-to-ceiling window with a mirror, and the result was pretty impressive using NB Pro from nearly a year ago. https://imgpb.com/ZXkiXV Locally hostable? For my money I'd argue Flux.2 Klein but Qwen-Edit still puts in the work.", "context": "", "use_case": "", "switched_from": "", "story": "Moebius: 0.2B image inpainting model with 10B-level performance"}, {"id": 777946416, "item_id": 48621651, "url": "https://news.ycombinator.com/item?id=48621651", "by": "vunderba", "date": "2026-06-21", "version": null, "stance": "recommend", "quote": "I remembering being absolutely shocked when the gpt-image series managed to pass the Labyrinth test", "excerpt": "Periodic motions coupled with \"whole image coherence\" is still very difficult even for non-SD based models (NB, Flux, etc.) I remembering being absolutely shocked when the gpt-image series managed to pass the Labyrinth test. https://genai-showdown.specr.net/#the-labyrinth", "context": "", "use_case": "AI image generation", "switched_from": "", "story": "The 100k Whys of AI"}, {"id": 774666368, "item_id": 48416648, "url": "https://news.ycombinator.com/item?id=48416648", "by": "jjcm", "date": "2026-06-05", "version": "2", "stance": "recommend", "quote": "I highly recommend doing image->code as the flow here.", "excerpt": "> Setup a Blog / Static site generator (Pelican), create a simple but stylish theme RE this one, I highly recommend doing image->code as the flow here. Codex's sites feature is doing this under the hood - it's rendering an image first with gpt-image-2, then building from it as a reference. You can use gpt-image-2 directly for this, though if I can plug my own stuff diffui.ai it's exactly what I made this for. It'll m", "context": "", "use_case": "generating design references for sites", "switched_from": "", "story": "Ask HN: What is your (AI) dev tech stack / workflow? (June 2026)"}, {"id": 771214177, "item_id": 48200886, "url": "https://news.ycombinator.com/item?id=48200886", "by": "vunderba", "date": "2026-05-19", "version": "2", "stance": "neutral", "quote": "", "excerpt": "Where did you get that info from? According to Google's own docs as well as my own image generation tests via the API, it supports up to 4K natively for gemini-3.1-flash-image-preview (aka NB2). It just defaults to 1K. But I didn't see anything in the docs stating that it's just a simple upscale for larger resolutions. https://ai.google.dev/gemini-api/docs/image-generation#generate-4k-images", "context": "", "use_case": "", "switched_from": "", "story": "Remove AI Watermarks"}, {"id": 771201426, "item_id": 48200089, "url": "https://news.ycombinator.com/item?id=48200089", "by": "vunderba", "date": "2026-05-19", "version": "2", "stance": "recommend", "quote": "proprietary models like Gemini 3.1 Flash Image and GPT Image 2", "excerpt": "Definitely. I run an entire site built around a series of benchmarks that focus on prompts of increasingly difficult complexity with a focus on adherence, and even the state-of-the-art local models are probably only about thirty percent as good as proprietary models like Gemini 3.1 Flash Image and GPT Image 2. Comparing Qwen-Image, Flux.2, ZiT, NB2, and gpt-image-2 https://genai-showdown.specr.net/?models=qi,nbp3,f2d", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool"}, {"id": 770546305, "item_id": 48159144, "url": "https://news.ycombinator.com/item?id=48159144", "by": "echelon", "date": "2026-05-16", "version": "2", "stance": "neutral", "quote": "Use it with 3D and a GPT Image 2 or Nano Banana Pro pass.", "excerpt": "> Outside trained images or objects, Hunyuan falls apart pretty hard. It's not too bad. It's fantastic for blocking out scenes for upscaling and conversion to video. Use it with 3D and a GPT Image 2 or Nano Banana Pro pass. Eg. https://getartcraft.com/media/m_xxtxfsdq3m5fwg5b8hjvkwxp4qybey https://getartcraft.com/media/m_gb4qexzqz1grskhn28f235fdtxfkvv https://getartcraft.com/media/m_p6wh750n0gc1s4a0ts88kp5pvnvned (Al", "context": "", "use_case": "", "switched_from": "", "story": "Image-blaster: Creates 3D environments, SFX, and meshes from a single image"}, {"id": 770195010, "item_id": 48137188, "url": "https://news.ycombinator.com/item?id=48137188", "by": "bilater", "date": "2026-05-14", "version": "2", "stance": "recommend", "quote": "leveraging GPT-Image-2 for creating a comic book", "excerpt": "/goals on both codex and claude is insane. I gave Claude a full end-to-end task of creating a graphic novel. I've been thinking about this idea of leveraging GPT-Image-2 for creating a comic book for a while, but didn't have the time to do it, so I said screw it, let's see if the models can figure it out. I gave it full autonomy in choosing the story, artwork, and creating the end product. Here is a summary of the st", "context": "", "use_case": "generating comic page images", "switched_from": "", "story": "I let Claude autonomously create a graphic novel"}, {"id": 770064226, "item_id": 48129014, "url": "https://news.ycombinator.com/item?id=48129014", "by": "pycassa", "date": "2026-05-13", "version": "2", "stance": "neutral", "quote": "But I will give GPT-Image-2 a try.", "excerpt": "Thank you so much for your suggestion regarding UI design. As my main expertise is not this, I need some tool to depend on to ground my projects somehow. Even though stitch by google and claude design are not perfect, they give me some starting point. And then, after building the actual working project, will iterate until I like the look of it. This is how I'm using these right now. I can't even itearte on these desi", "context": "", "use_case": "", "switched_from": "", "story": "Tell HN: Dont use Claude Design, lost access to my projects after unsubscribing"}, {"id": 769394610, "item_id": 48087163, "url": "https://news.ycombinator.com/item?id=48087163", "by": "jjcm", "date": "2026-05-10", "version": "2", "stance": "recommend", "quote": "the gpt-image-2 drop really legitized the bet", "excerpt": "I'm working on a diffusion-powered UI design tool. My short term goal is to make AI-designed UI not look like Tailwind. My long-term goal is to be Figma, but powered by diffusion. https://diffui.ai I quit Figma about 4mo ago to start working on this, and the gpt-image-2 drop really legitized the bet. I recently release Brands for diffui, which let you establish a design system and consistently generate with it. I mad", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: What Are You Working On? (May 2026)"}, {"id": 769216080, "item_id": 48076005, "url": "https://news.ycombinator.com/item?id=48076005", "by": "vunderba", "date": "2026-05-09", "version": "2", "stance": "warn", "quote": "It\u2019s not a great look when the biggest proponents and supposed experts still manage to make such a royal botch job of things.", "excerpt": "An even bigger issue to me is that people who most frequently tend to avail themselves of AI art also tend to be the least observant. So even when the AI-generated image could have been good or at least serviceable, they somehow still manage to bungle the whole process. I actually wrote a blog post about this, with concrete examples (and my manual fixes) taken from OpenAI when they demoed gpt-image-2 and from Karpath", "context": "", "use_case": "", "switched_from": "", "story": null}, {"id": 768631136, "item_id": 48039446, "url": "https://news.ycombinator.com/item?id=48039446", "by": "vunderba", "date": "2026-05-06", "version": "2", "stance": "recommend", "quote": "the results were surprisingly good as well", "excerpt": "Yeah, I agree, I think specificity is key. If you leave an LLM, or any AI model, to its own devices, it gravitates to the figurative \"mean of its training data\" which tends to be pretty sterile. I actually have some examples of my own comics that I used to put gpt-image-2 through its paces, and the results were surprisingly good as well: https://mordenstar.com/other/gpt-2-comics", "context": "", "use_case": "generating comics", "switched_from": "", "story": "Creating an animated manga with GPT Image 2.0 and Claude Code"}, {"id": 767021664, "item_id": 47938854, "url": "https://news.ycombinator.com/item?id=47938854", "by": "vunderba", "date": "2026-04-28", "version": "2", "stance": "recommend", "quote": "Gpt-Image in the editing benchmarks of my GenAI Showdown shot up from scoring 3/12 to 10/12.", "excerpt": "Nice article. From DALLE to the GPT Image series (1, 1.5, and now 2), it\u2019s crazy how much things have improved on the image model for OpenAI. Gpt-Image in the editing benchmarks of my GenAI Showdown shot up from scoring 3/12 to 10/12. The example on your site where it successfully creates/integrates valid QR codes and barcodes is pretty wild. I\u2019m curious how far this can go with extended thinking. From what I can tel", "context": "runs a GenAI image model benchmark", "use_case": "image generation and editing", "switched_from": "", "story": "Wild GPT-image-2 use cases"}, {"id": 766966112, "item_id": 47935382, "url": "https://news.ycombinator.com/item?id=47935382", "by": "vunderba", "date": "2026-04-28", "version": "2", "stance": "neutral", "quote": "", "excerpt": "Fun stuff. Mythology is steeped in these chimera-type creatures, where we mix and match body parts from different animals to create new ones Mr. Potato head style. There\u2019s a great Demetri Martin sketch about this, where he says: the horse becomes the majestic Pegasus, the lion becomes the griffin, and the hawk becomes... the double hawk. I've also used GenAI to visualize some of the creatures I've come up with as wel", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Photorealistic GPT Image 2 animal hybrids"}, {"id": 766850865, "item_id": 47928179, "url": "https://news.ycombinator.com/item?id=47928179", "by": "vunderba", "date": "2026-04-27", "version": null, "stance": "recommend", "quote": "ask a editing model such as Nano Banana or GPT Image to remove that leg.", "excerpt": "From the article: > If I truly wanted to use an LLM image generator to create this spider, I\u2019m sure it\u2019d be possible. Naturally. The easiest way to do it would be to generate a silhouette of a spider in the requested art deco style until you get one you like. Then, either quickly erase the leg manually or circle it with a red highlighter, and ask a editing model such as Nano Banana or GPT Image to remove that leg. Th", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.0 Still Can't Draw the Seven-Legged Spider I Want"}, {"id": 766011072, "item_id": 47875692, "url": "https://news.ycombinator.com/item?id=47875692", "by": "simonw", "date": "2026-04-23", "version": "2", "stance": "recommend", "quote": "here's gpt-image-2 as a recent example", "excerpt": "Honestly it never crossed my mind to waste some artist's time with this, but now that the joke \"benchmark\" has somehow reached orbital velocity maybe I should be thinking about it! I've run the prompt through dozens of dedicated image generation models so I've seen many versions of this that are better attempts than a text model spitting out SVG - here's gpt-image-2 as a recent example: https://chatgpt.com/share/69ea", "context": "", "use_case": "Image generation", "switched_from": "", "story": "Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model"}, {"id": 766003840, "item_id": 47875240, "url": "https://news.ycombinator.com/item?id=47875240", "by": "filearts", "date": "2026-04-23", "version": "2", "stance": "neutral", "quote": "Now look at GPT Image 2 and Nano Banana Pro; those original nits are mostly a thing of the past.", "excerpt": "Lot's of comments are nit-picking on how this demo misses the mark on specific technicalities of domains they're familiar with. But that's totally missing the point: this is a demo of what seems like a totally new interaction model that could not previously have existed. It's a small glimpse at a possible future where those nit-picky details get better. Remember when image models produced grotesque hands? Now look at", "context": "", "use_case": "image generation", "switched_from": "", "story": "Website streamed live directly from a model"}, {"id": 765763408, "item_id": 47860213, "url": "https://news.ycombinator.com/item?id=47860213", "by": "tezza", "date": "2026-04-22", "version": "2", "stance": "neutral", "quote": "gpt-image-2 has a lot more action, especially in the Apple Cart images.", "excerpt": "I've rushed out my standardised quality check images for gpt-image-2: https://generative-ai.review/2026/04/rush-openai-gpt-image-2-quality-settings/ I've done a series over all the OpenAI models. gpt-image-2 has a lot more action, especially in the Apple Cart images.", "context": "", "use_case": "AI image generation", "switched_from": "", "story": "OpenAI Livestream"}, {"id": 765752097, "item_id": 47859506, "url": "https://news.ycombinator.com/item?id=47859506", "by": "justani", "date": "2026-04-21", "version": "2", "stance": "compare", "quote": "One where Nano banana fails but gpt image 2 worked", "excerpt": "I have a few cases where nano banana fails all the time, even gpt image 2 is failing. A 3 * 3 cube made out of small cubes, with a small 2 * 2 cube removed from it - https://chatgpt.com/share/69e85df6-5840-83e8-b0e9-3701e9233244 Create a dot grid containing a rectangle covering 4 dots horizontally and 3 dots vertically - https://chatgpt.com/share/69e85e4b-252c-83e8-b25f-416984cf3040 One where Nano banana fails but gp", "context": "", "use_case": "image generation", "switched_from": "", "story": "OpenAI Livestream"}, {"id": 765742784, "item_id": 47858924, "url": "https://news.ycombinator.com/item?id=47858924", "by": "dzhiurgis", "date": "2026-04-21", "version": "2", "stance": "recommend", "quote": "", "excerpt": "Cost me < 1 cents - https://elsrc.com/elsrc/waldo/wojak.jpg p.s. aaaand that's soft launch my SaaS above, you can replace wojak.jpg with anything you want and it will paint that. It's basically appending to prompt defined by elsrc's dashboard. Hopefully a more sane way to manage genai content. Be gentle to my server, hn!", "context": "", "use_case": "image generation", "switched_from": "", "story": "OpenAI Livestream"}, {"id": 765730896, "item_id": 47858181, "url": "https://news.ycombinator.com/item?id=47858181", "by": "anshumankmr", "date": "2026-04-21", "version": "1", "stance": "neutral", "quote": "that is interesting cause I feel gpt-image-1 did have that feature.", "excerpt": "that is interesting cause I feel gpt-image-1 did have that feature. (source: https://chatgpt.com/share/69e83569-b334-8320-9fbf-01404d18dfa5)", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI Livestream"}, {"id": 765666624, "item_id": 47854164, "url": "https://news.ycombinator.com/item?id=47854164", "by": "vunderba", "date": "2026-04-21", "version": "1.5", "stance": "recommend", "quote": "gpt-image-1.5 was a big leap in visual quality", "excerpt": "OpenAI\u2019s gpt-image-1.5 and Google\u2019s NB2 have been pretty much neck and neck on my comparison site which focuses heavily on prompt adherence, with both hovering around a 70% success rate on the prompts for generative and editing capabilities. With the caveat being that Gemini has always had the edge in terms of visual fidelity. That being said, gpt-image-1.5 was a big leap in visual quality for OpenAI and eliminated m", "context": "", "use_case": "", "switched_from": "", "story": "ChatGPT Images 2.0"}, {"id": 755357937, "item_id": 47209871, "url": "https://news.ycombinator.com/item?id=47209871", "by": "vunderba", "date": "2026-03-01", "version": "1", "stance": "warn", "quote": "visually weak", "excerpt": "Yeah I think that it's part of the issue with a single \"squashed\" comparative metric. Some users are going to grade higher based on the overall visual fidelity and others are going to value following the prompt. For a point of reference, I run a pretty comprehensive image model comparison site heavily weighted in favor of prompt adherence. https://genai-showdown.specr.net EDIT: FWIW, I agree with your assessment. Ope", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Imagedojo.ai \u2013 Blind arena for Google, OpenAI, and xAI image generators"}, {"id": 755357938, "item_id": 47209871, "url": "https://news.ycombinator.com/item?id=47209871", "by": "vunderba", "date": "2026-03-01", "version": "1.5", "stance": "recommend", "quote": "until they finally pushed out gpt-image-1.5", "excerpt": "Yeah I think that it's part of the issue with a single \"squashed\" comparative metric. Some users are going to grade higher based on the overall visual fidelity and others are going to value following the prompt. For a point of reference, I run a pretty comprehensive image model comparison site heavily weighted in favor of prompt adherence. https://genai-showdown.specr.net EDIT: FWIW, I agree with your assessment. Ope", "context": "", "use_case": "", "switched_from": "", "story": "Show HN: Imagedojo.ai \u2013 Blind arena for Google, OpenAI, and xAI image generators"}, {"id": 751351876, "item_id": 46959492, "url": "https://news.ycombinator.com/item?id=46959492", "by": "cubefox", "date": "2026-02-10", "version": "1.5", "stance": "warn", "quote": "The first is very fake and the second is a strong improvement", "excerpt": "In your example, z-image and Nano Banana Pro look basically equally photorealistic to me. Perhaps the NBP image looks a bit more real because it resembles an unstaged smartphone shot with wide angle. Anyway, the difference is very small. I agree the lighting in Flux.2 Pro looks a bit off. But anyway, realistic environments like a street cafe are not suited to test for photorealism. You have to use somewhat more fanta", "context": "", "use_case": "", "switched_from": "", "story": "Qwen-Image-2.0: Professional infographics, exquisite photorealism"}, {"id": 749299376, "item_id": 46831211, "url": "https://news.ycombinator.com/item?id=46831211", "by": "echelon", "date": "2026-01-30", "version": "1.5", "stance": "recommend", "quote": "The best model I've found for this, that almost bakes in full ControlNet capability, is oddly gpt-image-1.5.", "excerpt": "It's funny to see as a joke, but you can go the other way with this too. Image editing models and LoRAs for \"previz-to-render upscaling\" workflows are actually incredibly useful. I was just writing about this (scroll about halfway down to the images of Sam Altman - though if you like that, do watch the second video): https://getartcraft.com/news/world-models-for-film The best model I've found for this, that almost ba", "context": "", "use_case": "Upscaling previz renders", "switched_from": "", "story": "Antirender: remove the glossy shine on architectural renderings"}, {"id": 741537280, "item_id": 46346080, "url": "https://news.ycombinator.com/item?id=46346080", "by": "vunderba", "date": "2025-12-21", "version": "1", "stance": "recommend", "quote": "GPT-image-1 (and gpt-image-1.5) are outliers in their ability to generate images with a proper alpha channel.", "excerpt": "It's a pretty well known problem in the generative image domain. GPT-image-1 (and gpt-image-1.5) are outliers in their ability to generate images with a proper alpha channel. You can still use other models (Flux, SDXL, NB, etc), you'll just need to request that the image be generated on a solid colored blackground. Then just run it through Rembg [1] or similar background removal tool. [1] - https://github.com/danielg", "context": "", "use_case": "Image generation with alpha", "switched_from": "", "story": "Why Can't Gemini Generate Images with Transparent Backgrounds?"}, {"id": 741537281, "item_id": 46346080, "url": "https://news.ycombinator.com/item?id=46346080", "by": "vunderba", "date": "2025-12-21", "version": "1.5", "stance": "recommend", "quote": "GPT-image-1 (and gpt-image-1.5) are outliers in their ability to generate images with a proper alpha channel.", "excerpt": "It's a pretty well known problem in the generative image domain. GPT-image-1 (and gpt-image-1.5) are outliers in their ability to generate images with a proper alpha channel. You can still use other models (Flux, SDXL, NB, etc), you'll just need to request that the image be generated on a solid colored blackground. Then just run it through Rembg [1] or similar background removal tool. [1] - https://github.com/danielg", "context": "", "use_case": "Image generation with alpha", "switched_from": "", "story": "Why Can't Gemini Generate Images with Transparent Backgrounds?"}, {"id": 740859088, "item_id": 46303693, "url": "https://news.ycombinator.com/item?id=46303693", "by": "encroach", "date": "2025-12-17", "version": "1.5", "stance": "recommend", "quote": "OAI's latest image model outperforms Google's in LMArena in both image generation and image editing.", "excerpt": "OAI's latest image model outperforms Google's in LMArena in both image generation and image editing. So even though some people may prefer nano banana pro in their own anecdotal tests, the average person prefers GPT image 1.5 in blind evaluations. https://lmarena.ai/leaderboard/text-to-image https://lmarena.ai/leaderboard/image-edit", "context": "", "use_case": "", "switched_from": "", "story": "Gemini 3 Flash: Frontier intelligence built for speed"}, {"id": 740825872, "item_id": 46301617, "url": "https://news.ycombinator.com/item?id=46301617", "by": "leumon", "date": "2025-12-17", "version": "1.5", "stance": "warn", "quote": "", "excerpt": "One other test you could add is generating a chessboard from a FEN. I was surprised to see NBP able to do that (however, it seems to only work with fewer pieces, after a certain amount it makes mistakes or even generates a completely wrong image) https://files.catbox.moe/uudsyt.png", "context": "", "use_case": "image generation and editing", "switched_from": "", "story": "GPT Image 1.5"}, {"id": 740759040, "item_id": 46297440, "url": "https://news.ycombinator.com/item?id=46297440", "by": "vunderba", "date": "2025-12-16", "version": "1.5", "stance": "compare", "quote": "GPT-Image 1.5 is clearly a step up as an editing model, but it performed worse in purely generative tasks", "excerpt": "I actually just finished running the Text-to-Image benchmark a few minutes ago. This matches my own testing as well. GPT-Image 1.5 is clearly a step up as an editing model, but it performed worse in purely generative tasks compared to its predecessor - dropping from 11 (out of 14) to 9. Comparing NB Pro, GPT Image 1, and GPT Image 1.5 https://genai-showdown.specr.net/?models=o4,nbp,g15", "context": "", "use_case": "", "switched_from": "", "story": "GPT Image 1.5"}, {"id": 740759041, "item_id": 46297440, "url": "https://news.ycombinator.com/item?id=46297440", "by": "vunderba", "date": "2025-12-16", "version": "1", "stance": "neutral", "quote": "dropping from 11 (out of 14) to 9", "excerpt": "I actually just finished running the Text-to-Image benchmark a few minutes ago. This matches my own testing as well. GPT-Image 1.5 is clearly a step up as an editing model, but it performed worse in purely generative tasks compared to its predecessor - dropping from 11 (out of 14) to 9. Comparing NB Pro, GPT Image 1, and GPT Image 1.5 https://genai-showdown.specr.net/?models=o4,nbp,g15", "context": "", "use_case": "", "switched_from": "", "story": "GPT Image 1.5"}, {"id": 740718945, "item_id": 46294934, "url": "https://news.ycombinator.com/item?id=46294934", "by": "vunderba", "date": "2025-12-16", "version": "1", "stance": "compare", "quote": "than gpt-image-1, doubling the previous score from 4/12 to 8/12", "excerpt": "Okay results are in for GenAI Showdown with the new gpt-image 1.5 model for the editing portions of the site! https://genai-showdown.specr.net/image-editing Conclusions - OpenAI has always had some of the strongest prompt understanding alongside the weakest image fidelity. This update goes some way towards addressing this weakness. - It's leagues better at making localized edits without altering the entire image's ae", "context": "", "use_case": "", "switched_from": "", "story": "GPT Image 1.5"}, {"id": 740446706, "item_id": 46277919, "url": "https://news.ycombinator.com/item?id=46277919", "by": "davidcann", "date": "2025-12-15", "version": "1", "stance": "neutral", "quote": "OpenAI\u2019s gpt-image-1 for the raw image generation", "excerpt": "I\u2019m working on Universymbols, which lets you restyle icons to match your app\u2019s icon set, like SF Symbols, Material Symbols, and many more. It uses a pipeline involving several AI models, including Nano Banana and OpenAI\u2019s gpt-image-1 for the raw image generation. https://universymbols.com/", "context": "", "use_case": "", "switched_from": "", "story": "Ask HN: What Are You Working On? (December 2025)"}], "latest_version": {"label": "GPT Image 2", "people": 17, "rec": 7, "warn": 2, "other": 8}}