全部 AI 动态

gemini 3.7 flash vs deepseek v4 pro 0813 vs muse spark 1.2 – on voxel city dioramas three models each built three crossy road-style 3d scenes – a construction site, a nyc intersection, a river with a drawbridge – as single self-contained html files the setup: @nousresearch's hermes agent cli on @openrouter, three.js skills preloaded, identical prompts per scene tasks: 1. construction site – tower crane on a working lift loop, paver laying fresh road, roller compacting it behind 2. nyc crossing – four-way intersection with a traffic light state machine, queuing cars, pedestrians crossing on the walk signal 3. river drawbridge – double-leaf bascule that lifts for tall boats, cars queuing at the barriers, animated water every scene: @threejs r185, box geometry only, a locked 20-color palette, four camera presets, and a day/night mode with bloom. one file, no build step, no assets models: @googledeepmind gemini 3.7 flash, @deepseek_ai v4 pro 0813, @aiatmeta muse spark 1.2 muse and gemini finished every scene in two to three minutes. deepseek took 15 to 41 minutes per scene - build time, all three scenes #1 gemini 3.7 flash – 6m 43s #2 muse spark 1.2 – 7m 20s #3 deepseek v4 pro – 91m 25s - total tokens #1 muse spark 1.2 – 440,279 #2 gemini 3.7 flash – 713,855 #3 deepseek v4 pro – 20,957,568 - total price #1 muse spark 1.2 – $0.53 #2 gemini 3.7 flash – $0.56 #3 deepseek v4 pro – $4.57 - agent calls across the three builds #1 muse spark 1.2 – 12 #2 gemini 3.7 flash – 18 #3 deepseek v4 pro – 143 observations: • muse won two of the three scenes on looks with the smallest files in the test – 887 to 1,042 lines against gemini's 1,934 to 2,377. cheapest, fastest to a good frame, and shortest turned out to be the same column • deepseek burned 20.96m tokens – 29x gemini, 48x muse – across 143 agent calls. prompt caching is the only reason that cost $4.57: the cache discount absorbed roughly $30 of resent context • gemini was the only model whose files needed zero fixes to render – and the only one whose night mode is cosmetic. the sky never darkens and one camera button does nothing. clean code for a scene it never looked at follow @thehypedotnews for 24/7 ai news, analysis and breakdowns

Grok Bot is the most powerful AI agent I've ever used. Learn to use it correctly, and you'll have agent swarms working for you 24/7. This master guide teaches you everything you need to know about the new Grok Bot (by @SpaceXAI ):

Grok 4.6 just ranked #2 in agentic U.S. administrative/regulatory legal research and the cost is insane • Grok 4.6 — 62.12% | $1.52/test • Claude Opus 5 — 65.15% | $6.58/test • GPT-5.6 Sol — 60.61% | $19.69/test Grok actually beats GPT-5.6 Sol on accuracy while costing roughly 13× less per test And it comes within just 3 percentage points of Opus 5, while Opus costs more than 4× as much This is exactly where Grok 4.6 is getting ridiculous It’s delivering frontier-level agentic performance at a fraction of the cost Grok is making high-end AI dramatically cheaper to actually use

1/6 🧵 MiniCPM-o 4.5 is a 9B open-source model built to see live video, hear incoming audio, and speak at the same time. In this video, we explain how full-duplex multimodal AI works—and show the LearnOpenCV experiments.

My AI startup made an LLM nearly 80% faster. An autonomous research system took MiniCPM5-1B from about 165 to 287 tokens/second on an RTX 3060. I show what worked, what failed, and why AI that optimizes AI could become a real business. Become an AI researcher with courses and 1-on-1 help from me: https://www.skool.com/become-ai-researcher-2669/about

Here's a comparison i never thought i would make Qwen-3.8-27b vs Gemini-3.7-flash Yes, you read it right! I don't know what to say, this is soo damn impressive. the difference is hugeee🤯 Take a look at this result 😍 (watch it in highest quality) Same prompt, both one shot on web-chat ui Qwen-3.8-27b produces consistent beautiful, visually stunning results. This is one of the best water surface simulations i have seen across every local models i have run and at the same time beating Gemini-3.7-flash result. And this is running on my 3090. And yes its Q5, not even f8.

🥇 We evaluated Grok 4.6 on MedAgentBench, a benchmark for agentic clinical EHR tasks, and it took the top spot. Grok 4.6 posts the highest pass@1 we've measured to date, moving ahead of the previous leader, GPT-5.6 Sol, to claim first place on the leaderboard. 🏥 Grok 4.6 averages ~95.9% pass@1 (avg of 3 runs), edging out the prior best (GPT-5.6-sol at ~94.7%) and improving on Grok 4.5 by roughly 2.5 points. On a benchmark where the model acts as an autonomous agent in a simulated EHR, calling FHIR APIs to complete clinical tasks across 10 task types, the result is also remarkably consistent run to run (95.3% to 96.3%). 📊 Leaderboards: https://medicalsphere.ai/benchmarks

I think Grok @Bot is a glimpse into the future of personal AI agents. Here's my new tutorial where I show you how to set up 5 useful bots: 1. An advisor to create and manage your bots 2. A YouTube researcher to find outlier videos 3. An X scout to find viral and funny tweets 4. A digital Marie Kondo to clean up your inbox and save money on paid subscriptions 5. A personal concierge to save money on trips I also tested a Gamer bot to see if Grok Bot can install and play classic games like Doom, Red Alert, and Commander Keen. Plus, I discuss the biggest barrier to Grok Bot adoption and whether it can replace ChatGPT as my daily driver. 📌 Watch now: https://youtu.be/MkVcHbviYOw

Artificial Analysis Intelligence Index puts Qwen3.8-27B at DeepSeek V4-Pro and GPT 5.6 Luna performance. This is the first time a local model has scored frontier model capability. We weren’t expecting this pace of local progress anywhere near this soon.

Uploading and logging into everything on OpenAI Finance is a better CPA and financial advisor than any human I’ve ever paid. Ditto contracts and lawyers, and results/scans and doctors. You only interface with humans out of the legacy capture and inertia of the offline world.

Agentic work is where @grok 4.6 lands hardest, taking the top spot on the Artificial Analysis Agentic Index at 59, tied with Claude Opus 5 Max. The index measures tool use, planning, autonomy and complex problem solving rather than single answers Grok 4.6 completes tasks in ~53 turns and ~0.5bn input tokens on average, against ~103 turns and ~2.0bn for Claude Opus 5 Max Cost of $0.84 per task, putting it on the intelligence versus cost per task Pareto frontier Enterprises buying agents pay per completed task, not per benchmark point. Turn efficiency is what determines whether a long-running workflow is affordable at volume. Two labs now sit at the top of this index with very different cost structures. Buyers get real choice on price for the first time in agentic deployment.

Couldn't resist the temptation to create this Music using Lyra 3.0 Pro on @pixverse CLI Video using @Hailuo_AI H3 on Hailuo platform Prompts below 👇

THE DRAGON RIDER — A CINEMATIC MOVIE TRAILER WITH PROMPTS What if Alexa Grey had her own Avengers-style superhero movie? I created this cinematic trailer using MiniMax H3 in PixVerse AI, mixing modern street fashion, kung fu, futuristic skateboarding and an ancient golden dragon spirit. And I wanted to push the lettering as hard as the action: 🔤 Giant 3D typography 🐉 Dragon energy wrapping around letters 🏙️ Neon billboards & city signage ⚡ Characters moving through typography 🎬 Cinematic camera moves & speed ramps 💥 Impact transitions and animated graphic cards Alexa isn't just fighting through the world. She's riding through it. 🐉 ALEXA GREY: THE DRAGON RIDER Made with MiniMax H3 in @pixverse #PixVerseCPP

Cool MiniMax-H3 use case unlocked for game sprites creation, here's the workflow: 1. one character design still, on a flat magenta background 2. one 5s H3 clip per move (walk, jump, punch, kick, block...) 3. cut each clip down to its pose extremes 4. key out the magenta, pack into a sprite atlas wire that to a state machine to make it playable (8 moves, ~$5)

Recently tried making AI videos locally for the first time using MiniMax H3 + ComfyUI, and I went through three different approaches: text-to-video, syncing to music beats, and image-to-video. 🎬 As an AI video newbie, my biggest takeaway is that the workflow itself isn't nearly as overwhelming as I thought - the real challenge is writing the prompts. Luckily, we can just use AI to solve that! ---- I’d drop my raw ideas straight to the AI, then run them through the official H3 Prompt Writing Skill to turn those thoughts into structured prompts. For example, with my first summer travel video, I broke it down into four scenes. I kept a similar structure across each scene and just swapped out the location details, running it iteratively and tweaking as I went until the output actually matched what was in my head. Thank goodness it's a local model-I could experiment endlessly without blowing through cash just to bridge the gap between "what I wanted" and "what the model understood." The absolute best surprise was that the whole process didn't take long at all. This model is insanely obedient and respects instructions really well, so I barely had to deal with gacha-style rerolls to get the exact shot I wanted. For the music beat-sync video, I wanted the visuals to shift on every kick drum or snare. The final result pretty much hit the nail on the head. Honestly, instead of just generating a video that looks "pretty good," I care way more about whether the model actually gets the rhythm and executes the pacing I set. And this time, it totally delivered. 🥁 Then for the seaside clip, there was a ton of scene description and creative direction, plus a 4-shot storyline-and the model managed to generate the visuals, the vibe, and the background audio all in one single pass. Genuinely blew me away. 🌊 Plus, it’s super beginner-friendly. You can just plug and play with the official workflow templates; literally all I had to think about was writing the prompts, choosing the resolution, and setting the duration. ------- Oh, quick side note on my local PC setup: * GPU: Consumer-grade RTX 4090 24GB * 20 steps for a 15-second video takes about 200+ seconds * Cranking it up to 768P brings it to 1,000+ seconds * Upscaling it to 2K using the official API takes around 6–8 minutes The speed is honestly pretty decent. I heard a bunch of speedup LoRAs came out recently that I haven't gotten around to testing yet, but those should make it run even faster! (Gotta love the power of open source.) ------- I’ll be packaging up all the workflows, scripts, and assets I used for this run and dropping them on GitHub for free!

Grok 4.6 just broke into the Top 3 on the Artificial Analysis Healthcare & Medical Index Grok 4.6 (high) scores 50 - just ONE point off the top score of 51 It outperforms GPT-5.6 Sol, Kimi K3 and Gemini 3.7 Flash This index measures how models actually handle healthcare and medical work: • Medical and health knowledge • Clinical reasoning • Patient documentation • Non-hallucination • Agentic workflows Healthcare is one of the highest-stakes domains you can test an AI on And Grok is now competing at the very top of it

Meet Spider-Girl！ 🎬 Made with MiniMax H3 on PixVerse @PixVerse_ project link 👇 #H3onPixVerse #PixVerse #HeroVerse

On one hand, these are obviously much worse "otter using wifi on an airplane" than any state-of-the AI text-to-video generation, it looks like something from 2022. On the other, it was done entirely offline on my computer using open AI video generation tools, a new capability.

BREAKING: Seedance 2.5 by @BytePlusGlobal ranks 1st in Multi-Image to Video on Design Arena with an Elo of 1400. Open-weight MiniMax H3 by @MiniMax_AI takes 2nd with an Elo of 1355, the top open source video model. Unlike standard image-to-video, this leaderboard gives models multiple reference images and compares their final video outputs. This tests how well models combine different visual inputs while preserving subjects, styles, and scene continuity - a critical capability for controlled video creation. Congratulations to both teams!

THE DRAGON RIDER — A CINEMATIC MOVIE TRAILER WITH PROMPTS What if Alexa Grey had her own Avengers-style superhero movie? I created this cinematic trailer using MiniMax H3 in PixVerse AI, mixing modern street fashion, kung fu, futuristic skateboarding and an ancient golden dragon spirit. And I wanted to push the lettering as hard as the action: 🔤 Giant 3D typography 🐉 Dragon energy wrapping around letters 🏙️ Neon billboards & city signage ⚡ Characters moving through typography 🎬 Cinematic camera moves & speed ramps 💥 Impact transitions and animated graphic cards Alexa isn't just fighting through the world. She's riding through it. 🐉 ALEXA GREY: THE DRAGON RIDER Made with MiniMax H3 in @PixVerse_ #PixVerseCPP

Here is how to enable a 1M-token context window in Codex for GPT-5.6 Sol. Even though we have tuned the context limit in Codex to be set optimally when it comes to performance and cost, this is a common ask, so here it is documented. A larger context window lets Codex retain more code, tool output, and conversation history before summarizing older material. You need a model that supports it. And GPT-5.6 Sol, for example, has a documented 1,050,000-token window. Open ~/.codex/config.toml and add or update these settings at the top level, before any [section] headers: ``` model = "gpt-5.6-sol" model_context_window = 1000000 model_auto_compact_token_limit = 900000 ``` The first setting selects the model. The second tells Codex to use a one-million-token context budget. The third starts automatic history compaction around 900,000 tokens, leaving some headroom. Restart Codex client and start a new session after saving. To try the configuration for a single CLI session without changing your defaults: ``` codex -m gpt-5.6-sol \ -c model_context_window=1000000 \ -c model_auto_compact_token_limit=900000 ``` Have fun, but also know that we tuned the default carefully!