MiniMax (official)@MiniMax_AI · 6月1日64It truly is 😎 #M3
译确实如此 😎 #M3
MiniMax (official)@MiniMax_AI · 6月1日77M3 live on @novita_labs 🔥
it's time to build (50% off the first week 👀)
译MiniMax M3 模型现已在 Novita AI 平台上线,并提供首周半价优惠。作为首个开源权重模型,它集成了前沿编码与智能体能力,在 SWE-Bench Pro 上得分 59.0%,Terminal Bench 2.1 上得分 66.0%,MCP Atlas 上得分 74.2%。该模型上下文窗口最高可达 1M tokens,由 MiniMax Sparse Attention 技术支持,并从一开始即支持原生多模态,可处理文本与视觉理解任务。Novita AI 作为其 Day-0 API 发布合作伙伴,为开发者提供接入服务。
Berryxia.AI@berryxia · 6月1日71刚刚在Hugging Face刷新模型时,看到KwaiKeye放出了Keye VL 2.0-30B-A3B。
这个多模态模型总参数30B,活跃参数只有3B,Apache 2.0完全开源。
它直接用DeepSeek Sparse Attention实现了256K上下文。
最有意思的是视频理解部分的表现。
你喂给它的帧数越多,模型准确率反而稳步上升。
这和我们以前觉得长视频容易让模型迷失的直觉完全相反。
它在多个长视频基准上已经和Qwen3 VL、Gemini 3 Flash打成平手。
以前大家总觉得多模态模型要么上下文够长,要么理解够深,二者很难兼得。
现在KwaiKeye把稀疏注意力真正落地,把这两件事同时推到一个新水平。
实际效果如何,后面看看真实case册书。
译KwaiKeye开源了多模态大模型Keye VL 2.0-30B-A3B,采用Apache 2.0许可。该模型总参数为30B,但仅激活3B参数。其核心亮点是通过DeepSeek稀疏注意力技术实现了256K的上下文长度。该模型的视频理解能力表现出一个反直觉的特性:喂入的帧数越多,其准确率反而持续上升。在基准测试中,其表现已与Qwen3 VL、Gemini 3 Flash等模型相当。
歸藏(guizang.ai)@op7418 · 6月1日76MiniMax 这次终于发布了他们的大版本号模型升级 MiniMax M3。
主要是三个核心能力:
1. 标配了 1M 的超长上下文
2. 采用新的 MSA(MoE with Segment-wise Attention)稀疏注意力架构
3. 从训练起就融合了原生多模态能力,包括强化了文本、图片、视频和桌面操作这些多模态的训练。
MSA 超上下文在算力和速度上更可落地,在 100 万上下文下每 token 计算量只有上一代的约 1/20。。
所以他们这一次的价格也比较给力,Token Plan 更新了新的价格。
在 API 上,小于512k 的 API 现在有为期 7 天的限时五折。
译MiniMax 发布了其大版本号模型升级 MiniMax M3。该模型标配 1M 超长上下文,采用新的 MSA(MoE with Segment-wise Attention)稀疏注意力架构,在 100 万上下文下每 token 计算量降至约上一代的 1/20。M3 从训练起即融合了原生多模态能力。在基准测试中,其取得了 SWE-Bench Pro 59.0%、Terminal Bench 2.1 66.0%、MCP Atlas 74.2% 等成绩。此外,其 API 推出小于 512k 调用的限时七天五折优惠。模型权重与技术报告预计约 10 天后发布。
MiniMax (official)@MiniMax_AI · 6月1日74This is hype 🔥
M3 live on @OrcaRouter on day one
go run it (50% off the first week 👀)
译这很火🔥
M3 首日上线 @OrcaRouter
快去跑起来(首周五折👀)
🚨 AI News | TestingCatalog@testingcatalog · 6月1日80MINIMAX 🔥: A new open-weights model, MiniMax M3, has been released to the public on APIs and MiniMax Agent.
MiniMax M3 scores 59% on SWE Bench Pro (on par with GPT-5.5), supports a 1M context window via MiniMax Sparse Attention, and is natively multimodal.
MiniMax Agent Updates 👀
> Meet M3: Our most intelligent and responsive model designed to handle any task.
> Persistent Memory: Your Agent remembers what you've shared, so you never have to repeat yourself.
> Evolving Skills: It learns as you collaborate, turning complex tasks into skills made just for you.
> Unified Billing: Fully integrated with Token Plan for a smoother, more consistent experience.
译MiniMax发布了新开源权重模型M3,现已通过API和MiniMax Agent提供服务。该模型在SWE-Bench Pro上得分59.0%,在Terminal Bench 2.1上得分66.0%,并支持高达1M的上下文窗口。同时,MiniMax Agent更新了持久记忆与进化技能等能力。此外,MiniMax Code也已发布,模型权重与技术报告将在约10天后公开。
Artificial Analysis@ArtificialAnlys · 6月1日81NVIDIA just announced the release of Nemotron 3 Ultra in Jensen Huang's Computex keynote: at 550B parameters (55B active), this is the largest Nemotron 3 model to date, and it is the most intelligent US open weights model
We partnered with @nvidia to evaluate this model for intelligence and speed - these figures use the model’s BF16 weights, but as with Nemotron 3 Super the model will be made available in NVFP4 quantization as well for higher inference performance.
➤ New leader for US open weights intelligence: Nemotron 3 Ultra scores 48 on the Artificial Analysis Intelligence Index. This is well ahead of the next strongest US open weights models, Gemma 4 31B (39), Nemotron 3 Super (36) and gpt-oss-120b (33), but behind the Chinese-led open weights frontier (Kimi K2.6 at 54).
➤ Leading speed for its intelligence: on a pre-release @DeepInfra endpoint, Nemotron 3 Ultra served over 300 tokens per second. Peer models in its size class from China-based labs such as DeepSeek and Moonshot (Kimi) are generally served at speeds of 50-100 tokens per second in the market today. gpt-oss-120b is served at speeds similar to this level, but with significantly lower intelligence.
➤ Largest Nemotron 3 model so far: at approximately 550 billion total parameters and 90% sparsity, Nemotron 3 Ultra is significantly larger than its siblings and is the largest recent US open weights model release
We’ll be sharing additional analysis and full benchmarks at release.
译NVIDIA在Computex上发布了Nemotron 3 Ultra,总参数达550B(激活参数55B),是目前最大的Nemotron 3模型。该模型在美国开放权重模型中智能性最强,在Artificial Analysis Intelligence Index评测中得分为48,超越了Gemma 4 31B(39分),但仍落后于月之暗面(Kimi)的K2.6(54分)。在推理速度方面,其在预发布端点上超过了300 tokens/s,远高于同级别中国模型通常的50-100 tokens/s。该模型将提供BF16权重及NVFP4量化版本以提升推理性能。
MiniMax (official)@MiniMax_AI · 6月1日69@CreaoAI moving fast 🔥 M3's live on day one, go try it
译@CreaoAI 行动迅速 🔥 M3在第一天就上线了,快去试试
[引用 @CreaoAI]:MiniMax M3现已在CREAO上线。
采用稀疏注意力推理,在长上下文下解码速度最高提升15.6倍,专为需要处理海量代码库、文档和转录文本而不减速的智能体打造。
从模型下拉菜单中选择M3即可运行。⚡
MiniMax (official)@MiniMax_AI · 6月1日72@cline put it well 🔥 sparse attention cutting compute to 1/20th the previous gen.
Try it for free on Cline right now
译@cline 说得好 🔥 稀疏注意力将计算量降至前代的1/20。
现在可以在 Cline 上免费试用。
MiniMax (official)@MiniMax_AI · 6月1日78let's gooo 🔥 M3 + @visionagents_ai for real-time voice & video
go build something hype!
译出发吧🔥 M3 + @visionagents_ai 实现实时语音与视频
去构建些令人兴奋的东西吧!
[引用 @visionagents_ai]:祝贺 @MiniMax_AI 团队发布 M3!
👉 一款前沿级开源权重模型
👉 100万 token 上下文窗口
👉 原生多模态(图像与视频)
MiniMax (official)@MiniMax_AI · 6月1日76M3 on @OpenRouter same day we dropped it 🔥.
1M context, frontier coding + agentic, native multimodal. 50% off the first week.
译M3在我们发布当天就登陆了OpenRouter 🔥。
100万token上下文,前沿编码+智能体能力,原生多模态。首周半价。
MiniMax (official)@MiniMax_AI · 6月1日71M3 on @AskVenice, available anonymously 🔥
open-weight, frontier coding + agentic, 1M context, native multimodal.
Live on day one
译M3现已在@AskVenice上线,支持匿名使用🔥
开源权重,前沿编码与智能体能力,1M上下文,原生多模态。
首日即上线
MiniMax (official)@MiniMax_AI · 6月1日74Happy Children’s Day with a huge M3 drop!
译儿童节快乐,M3模型重磅发布!
OpenRouter@OpenRouter · 6月1日79MiniMax-M3 is live on OpenRouter!
A frontier-class open-weight model that combines a 1M-token context window, frontier coding and agentic performance, and native multimodality (image & video) in one model.
译MiniMax-M3现已在OpenRouter上线!
一款前沿级开源权重模型,结合了1M-token上下文窗口、前沿的编码和智能体性能,以及原生多模态(图像与视频)能力。
elvis@omarsar0 · 6月1日46MiniMax M3 imminent.
Will be doing deep testing with it on my own coding agent and harness.
Review coming soon.
译MiniMax M3 即将发布。
我将用自己的编码智能体和测试框架对其进行深度测试。
评测即将推出。
MiniMax (official)@MiniMax_AI · 6月1日79Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities
- Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas
- MiniMax Sparse Attention scales context to 1M
- Natively Multimodal from Step Zero
API: http://platform.minimax.io
Token Plan: https://platform.minimax.io/subscribe/token-plan
🚀New! MiniMax Code: http://code.minimax.io
Weights & Tech Report in ~10 Days
译介绍 MiniMax M3:首个融合三大前沿能力的开源权重模型
- 编码与智能体前沿:59.0% SWE-Bench Pro,66.0% Terminal Bench 2.1,34.8% SWE-fficiency,28.8% KernelBench Hard,74.2% MCP Atlas
- MiniMax Sparse Attention 将上下文窗口扩展至 1M
- 从零开始原生多模态
API:http://platform.minimax.io
Token 计划:https://platform.minimax.io/subscribe/token-plan
🚀新!MiniMax Code:http://code.minimax.io
权重与技术报告将在约 10 天内发布
OpenCode@opencode · 6月1日63MiniMax M3 will be launching soon
You can try it right now in OpenCode
For free
译MiniMax M3即将发布
你现在就可以在OpenCode中免费试用
Artificial Analysis@ArtificialAnlys · 6月1日74HiDream-O1-Image-Dev-2604 debuts as the leading open weights Text to Image model in the Artificial Analysis Image Arena, with the base HiDream-O1-Image and HiDream-O1-Image-Dev also available open weights but landing lower on the leaderboard
@HiDream_AI's O1-Image family spans three models: the 8B HiDream-O1-Image, its distilled HiDream-O1-Image-Dev, and HiDream-O1-Image-Dev-2604, a fine-tune of Dev with a prompt-enhancement pipeline, previously listed pseudonymously as Peanut. The base and Dev models accept text plus up to 10 image inputs, spanning generation and instruction-based image editing.
On the Artificial Analysis Text to Image Arena, HiDream-O1-Image-Dev-2604 leads all open weights models, delivering quality similar to proprietary models like ByteDance's Seedream 4.0 and Black Forest Labs' FLUX.2 [max]. In Image Editing, HiDream-O1-Image is the second-highest open weights model, behind only Tencent's HunyuanImage 3.0 Instruct.
Weights and the full inference pipeline (including HiDream's prompt refiner used during evaluation for HiDream-O1-Image-Dev-2604) are open-source on Hugging Face and GitHub under the MIT license.
HiDream-O1-Image and HiDream-O1-Image-Dev are also available across third-party API providers including Fal, priced on Fal at $10/1k images and $5/1k images respectively.
Congratulations to @HiDream_ai on the releases!
See below for comparisons between the HiDream-O1-Image family and other leading models in the Artificial Analysis Image Arena 🧵
译HiDream发布O1-Image系列文生图模型,包含8B参数的HiDream-O1-Image、其蒸馏版本HiDream-O1-Image-Dev,以及基于Dev微调并集成提示增强管线的HiDream-O1-Image-Dev-2604。在Artificial Analysis Text to Image Arena榜单上,Dev-2604版本在所有开源权重模型中排名第一,生成质量接近Seedream 4.0和FLUX.2 [max]等闭源模型。在图像编辑任务中,HiDream-O1-Image是排名第二高的开源模型,仅次于腾讯的HunyuanImage 3.0 Instruct。所有模型的权重及完整推理管线均以MIT许可证开源。HiDream-O1-Image与HiDream-O1-Image-Dev也通过Fal等第三方API提供,价格分别为$10/1k images和$5/1k images。
🚨 AI News | TestingCatalog@testingcatalog · 5月31日74Grok Imagine Video 1.5 Preview is now available on Grok APIs and has also landed in the top 1 spot on Video Arena.
Have you tested it already? 👀
译Grok Imagine Video 1.5 Preview 现已上线 Grok API,并在 Video Arena 排行榜上位列第一。
你已经测试过了吗?👀
DogeDesigner@cb_doge · 5月31日70NEW: Grok Imagine Video 1.5 Preview just hit #1 in the Image-to-Video Benchmark on Video Arena.
A massive +52 point jump over the previous Grok Imagine Video model, beating Seedance 2.0, HappyHorse, and Veo 3.1.
xAI is moving fast. 🚀
译新消息:Grok Imagine Video 1.5 Preview 刚刚在 Video Arena 的图生视频基准测试中排名第一。
相比之前的 Grok Imagine Video 模型,分数大幅提升了 52 分,超越了 Seedance 2.0、HappyHorse 和 Veo 3.1。
xAI 进展迅速。🚀
🚨 AI News | TestingCatalog@testingcatalog · 5月31日62BUILD 🔥: Microsoft is preparing new image and voice models for the announcement on June 2.
> MAI Voice 2, a multilingual model supporting 15 news languages and a wider range of emotional spectrum (check voice samples in the article)
> MAI Transcribe 1.5, a new model for speech-to-text use cases.
> MAI Image 2.5, already announced last week, is now available on LM Arena in preview. Compared to MAI Image 2, it supports file uploads and can be used for image editing.
译BUILD 🔥: 微软正为 6 月 2 日的发布会准备新的图像和语音模型。
> MAI Voice 2,一个支持 15 种新闻语言和更广泛情感光谱的多语言模型(请在文章中查看语音样本)
> MAI Transcribe 1.5,一个用于语音转文本用例的新模型。
> MAI Image 2.5,已于上周公布,现已在 LM Arena 上以预览版形式提供。与 MAI Image 2 相比,它支持文件上传,并可用于图像编辑。
StepFun@StepFun_ai · 5月30日67Step 3.7 Flash, free for 30 days for Hermes Agent users. What could possibly go wrong? 🍿
Thanks @NousResearch for making it happen. Can’t wait to see what Hermes users build!
译Step 3.7 Flash,Hermes Agent 用户可免费使用 30 天。还能出什么问题?🍿
感谢 @NousResearch 促成此事。迫不及待想看 Hermes 用户们会构建出什么!
StepFun@StepFun_ai · 5月30日55Curious to hear your verdict once it’s loaded. 👀
译等加载完想听听你的看法。👀
Google AI Developers@googleaidevs · 5月30日71ICYMI: Nano Banana Pro [gemini-3-pro-image] and Nano Banana 2 [gemini-3.1-flash-image] are now GA and ready for production via the Gemini API.
Check out these great community examples to see the capabilities of both models in action 🧵
译ICYMI:Nano Banana Pro [gemini-3-pro-image] 和 Nano Banana 2 [gemini-3.1-flash-image] 现已正式发布,可通过 Gemini API 投入生产使用。
查看这些优秀的社区示例,了解两个模型的实际能力 🧵
StepFun@StepFun_ai · 5月30日62Small is a feature. 😌
译小即是美。😌
Berryxia.AI@berryxia · 5月30日70Grok-building-0.1 支持使用𝕏 AI的 API了。
看了一下价格,输入:百万Token花费 1美金左右。
输出:百万Token花费 2美金左右。
译Grok-build-0.1 现已支持使用𝕏 AI的 API了。
看了一下价格,输入:百万Token花费 1美金左右。
输出:百万Token花费 2美金左右。
🚨 AI News | TestingCatalog@testingcatalog · 5月30日78xAI released grok-build-0.1, the model powering Grok Build CLI, on the API console in public beta.
> $1/m input and
> $2/m output
译xAI在API控制台公开发布了grok-build-0.1模型,该模型为Grok Build CLI提供支持。
> 输入价格为每百万token $1,输出价格为每百万token $2。
Rohan Paul@rohanpaul_ai · 5月30日64Today’s edition of my newsletter just went out.
🔗 https://www.rohan-paul.com/p/anthropic-releases-claude-opus-48
🗞️ Anthropic releases Claude Opus 4.8 on the same day as its $965B valuation round.
🗞️ KogAI just achieved 3,000 tokens/s on 8× AMD MI300X GPUs and 2,100 on 8× NVIDIA H200 (FP16, no speculative decoding) with a 2B model.
🗞️ Video to Watch: Transformer vs Post-Transformer, argued by leading researchers, inside a real physical boxing ring.
🗞️ Anthropic secures a massive post-money valuation of $965B after raising $65 B.
🗞️ Datacurve launches DeepSWE, a tougher coding benchmark made to show where leading models truly separate.
🗞️ OpenAI and Thrive just built a self-improving tax agent with up to 97% accuracy.
译本期简报要点如下:Anthropic发布了Claude Opus 4.8模型,并宣布完成650亿美元融资,投后估值达到9650亿美元。KogAI展示了其在特定硬件上的性能:使用8块AMD MI300X GPU时处理速度达3000 tokens/s,使用8块NVIDIA H200 GPU时达2100 tokens/s(FP16精度,无推测解码),模型参数为20亿。此外,Datacurve推出了更具挑战性的编程基准测试DeepSWE,旨在更清晰地评估顶尖模型的性能差异。
Greg Brockman@gdb · 5月30日76OpenAI for realtime translation — speak in any of 70+ input languages and translate into 13 output ones:
译OpenAI 实时翻译功能——使用70多种输入语言说话,翻译成13种输出语言:
gpt-realtime-translate 接收任意语言的语音输入,并输出目标语言的语音。
大语言模型很棒,但特定用例需要专用模型。
我们正在智能眼镜上运行此功能。
Chubby♨️@kimmonismus · 5月30日46Despite Opus 4.8, I somehow found myself sticking with GPT 5.5 and Codex. It is clear that GPT-5.6 is on the way; the fact that a new checkpoint has been reached pleases me all the more.
I have high expectations for the new model. What a year for OpenAI.
译尽管有 Opus 4.8,我发现自己仍在使用 GPT 5.5 和 Codex。显然 GPT-5.6 正在路上;一个新的检查点已经达成,这让我更加高兴。
我对新模型抱有很高的期望。OpenAI 这一年真是精彩。
Berryxia.AI@berryxia · 5月30日69最近开发了一个OCR的 工具,疯狂给干法律的客户案例!
效果非常好,很合适~
但也遇到有些错乱和不好的结果
金融合同、法律文件、研究报告、历史档案,这些东西里公式、表格、印章、稀有字符混在一起,传统工具经常认错或者直接漏掉,导致后续LLM输出质量直接拉低。
今天PaddlePaddle把PaddleOCR-VL 1.6正式发布了。
它在OmniDocBench上刷到96.33%的SOTA,把开源和商业方案同时甩在身后。
表格识别、经典文本、稀有字符都有明显提升,印章检测、图表理解也更稳。
最实用的是,它和1.5版本完全架构兼容,零迁移成本,拿来就能用。
以前大家总觉得RAG的瓶颈在模型参数或者检索算法,现在看,真正决定上限的往往是输入数据的干净程度。
这份高精度解析能力,直接把文档智能喂给LLM的门槛又往下拉了一大截。
译飞桨发布了PaddleOCR-VL 1.6版本。该版本在OmniDocBench评测基准上取得了96.33%的新SOTA成绩,在该榜单及Real5-OmniDocBench上均排名第一。在表格、经典文本和稀有字符识别能力上均有显著提升,并增强了印章检测与图表理解能力。该版本与1.5版本架构完全兼容,实现了零迁移成本,方便直接部署使用,旨在为大语言模型和检索增强生成等系统提供更高质量的输入数据。
xAI@xai · 5月30日73grok-build-0.1 is now available via the xAI API in public beta.
This is the same model that powers the Grok Build CLI and excels at agentic coding.
Priced at $1/m input and $2/m output, it’s extremely cost effective, intelligent, and fast.
译grok-build-0.1 现已通过 xAI API 提供公开测试版。
这是驱动 Grok Build CLI 的同一模型,擅长智能体编码。
定价为输入 $1/m,输出 $2/m,极具成本效益、智能且快速。
StepFun@StepFun_ai · 5月29日70Appreciate it 🙏 Step 3.7 Flash was built for agentic workflows — curious to see what @NousResearch users build with it on Hermes Agent.
译感谢 🙏 Step 3.7 Flash专为智能体工作流打造——期待看到@NousResearch用户用它在Hermes Agent上构建什么。
StepFun@StepFun_ai · 5月29日68Step 3.7 Flash is now live in @kilocode !
Open weights, ready to run. Appreciate the quick integration 🙏
译Step 3.7 Flash 现已在 @kilocode 上线!
开放权重,随时可运行。感谢快速集成 🙏
Greg Brockman@gdb · 5月29日69new 5.5 instant model in chatgpt:
译ChatGPT 新增 GPT-5.5 instant 模型:
我们今天发布了 GPT-5.5 instant 的新版本。之前的模型过于“子弹化”。新版本在其他一些重要维度上有所改进:谄媚性、事实性和多语言性能。希望你喜欢!我们始终对反馈感兴趣。
StepFun@StepFun_ai · 5月29日71Step 3.7 Flash now showing up on @ZenMuxAI —
nice to see it plugged into more model stacks!
译阶跃星辰(Step Fun)的视觉语言模型Step 3.7 Flash已在ZenMux平台上线。该模型采用稀疏MoE架构,专为智能体、编程、搜索、多模态及长上下文工作流设计。其核心性能包括:400 TPS推理速度、约110亿激活参数、256K上下文窗口及3个推理级别。该模型能够理解UI、图表、文档和图像以编写代码或调用工具,并擅长深度网络与视觉搜索,在τ²-bench上跨难度级别取得98%+的成绩。它兼容Claude Code、MCP风格工作流等,并可本地部署于Mac Studio M4 Max、DGX Spark等硬件。
StepFun@StepFun_ai · 5月29日71Step 3.7 Flash is now live on @OpenRouter !
should make it easier to plug into existing workflows.
译Step 3.7 Flash现已在 @OpenRouter 上线!
这应该能让它更容易地融入现有工作流。
StepFun@StepFun_ai · 5月29日72Excited to see Step 3.7 Flash available on @ModelScope2022 🚀
Can’t wait to see what builders create with it!
译阶跃星辰发布的多模态模型 Step 3.7 Flash 已在 ModelScope 平台上线。该模型采用 198B 总参数的 MoE 架构,每个 token 激活 11B 参数,推理速度最高达 400 tok/s,支持 256K 上下文窗口,并提供低、中、高三个推理级别以平衡速度与效果。其在 ClawEval-1.1 榜单位列第一(67.1分),在 SWE-bench Pro 上排名第二(56.3分)。模型具备原生多模态能力,由语言骨干与视觉编码器组成,原生支持解析密集UI、图表及财报。该模型采用 Apache 2.0 协议开源,并兼容 vLLM 等多种推理框架。
OpenRouter@OpenRouter · 5月29日69Step 3.7 Flash from @StepFun_ai is live on OpenRouter.
A multimodal (image/video/text) MoE that activates just 11B of 196B params. Tuned for coding, agentic workflows, and structured outputs, with selectable reasoning levels to trade off speed, cost, and depth.
译阶跃星辰的Step 3.7 Flash已在OpenRouter上线。
这是一个多模态(图像/视频/文本)MoE模型,196B参数中仅激活11B。针对编码、智能体工作流和结构化输出进行了调优,并提供可选的推理级别,以权衡速度、成本和深度。
Berryxia.AI@berryxia · 5月29日66制作飞机的提示词👇🏻
PROMPT:
Write a complete single-file HTML document using Three.js (include via CDN) that renders a highly detailed Boeing 747-400 aircraft.
Rules (must follow strictly):
- Use ONLY built-in Three.js primitive geometries (BoxGeometry, CylinderGeometry, ConeGeometry, SphereGeometry, etc.). NO external model loaders, no GLTFLoader, no imported assets.
- Pay extreme attention to accurate real-world proportions, swept wing angle (~35°), engine positions (4 engines under wings, correct size and spacing), tail configuration (horizontal + vertical stabilizer), cockpit windows, landing gear (retractable), fuselage curvature, winglets, etc.
- Make the model instantly recognizable as a Boeing 747 from front, side, top, and 3/4 views.
- Add realistic lighting (directional + ambient + point lights), soft shadows, OrbitControls for rotation, and a simple background (sky gradient or grid).
- Include a short animation (e.g. slow rotation or landing gear deployment) or make it fully interactive.
- Output the complete, ready-to-run HTML code in one block.
Start coding now.
译Anthropic发布Claude Opus 4.8,其复杂空间推理与代码生成能力受到关注。有用户使用其测试生成一架高细节波音747-400的Three.js模型,要求仅使用内置几何体,生成完整的单文件HTML。Claude Opus 4.8一次生成了可运行代码,模型具有后掠机翼约35度、四发动机、可收放起落架等细节,比例严谨。ZenMux平台现已支持该模型的API调用与免费体验。据称,Claude Opus 4.8在SWE-bench、Terminal-Bench、Agentic Coding等榜单排名第一。