全部 AI 动态

Grok Build, with the Browser Use CLI, outperforms Codex and Claude Code at browser tasks

I solved 6 open Erdős problems in 5 days, using @OpenAI GPT-5.6 Sol. I have a math background, but the Codex workflow I used does not require deep mathematical knowledge. Here’s exactly how I approached it, including my prompts 🧵

@OpenAI didn’t publish GPT-Live's end-to-end latency. So Agora Media Lab measured it on an iPhone – with repeatable speech, dual-track waveform recordings, and controlled network impairment. https://www.agora.io/en/blog/openai-didnt-publish-gpt-lives-latency-so-we-measured-it/ #RealTimeAI #VoiceAI

Meta's Muse Spark 1.1 beat Gemini 3.6 at making 3D arcade games for 3x cheaper! We gave two models the same task: build three 3D tabletop games that play themselves Prompts: - An 8-ball pool break scattering the rack - An air hockey rally that ends in a goal - A foosball table where the rods play the ball Outputs: Muse Spark 1.1: 31.3K tokens, $0.026 Gemini 3.6: 28.8K tokens, $0.073 Muse Spark played all three tables clean with far more detail than Gemini, which barely rendered its backgrounds. Its air hockey glitched and its foosball lagged into a buggy still picture. Muse Spark 1.1 ran the full set perfectly for 3x less

@kagigz I love using sites with chatgpt work on mobile, being able to host an app I made while talking to codex on the go is awesome.

My book, Reinforcement Learning from Human Feedback is done! This is the book I wish I had when learning to fine-tune, align, & now post-train models since ChatGPT. The resource has been built by me finding time to study and document the fundamentals on nights and weekends since 2024. Transferring as much of the intuitions of building Olmo as I possibly can in the book format. The book is launching with an over 10 hour, full course with slidedecks, functional code for the training chapters, an example model completions library, and of course the free online web version. Physical orders from Manning will ship in 1-2 weeks, and Amazon a week or so after. Thanks for your support!

Hy3 by Tencent is #5 in Agent Arena for open-weight models (#25 overall)! It also ranks as the #2 open model in the Frontend Code Arena (#16 overall)! In Agent Arena: Hy3 lands at #25 overall (net -2.2%). Hy3 has strengths in tool-use (recovering well from CLI/bash errors, +2.6% and #25), its biggest weakness is steerability as it struggles to course-correct when users push back, coming in at -7.1% (#30). Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models get web search, filesystem, and terminal tools to complete complex workflows: writing code, creating slide decks, researching the web, building apps, and analyzing documents. We use causal tracing methodology to measure a model's net improvement, which indicates how much it improves outcomes relative to the average model. Below we break down how Hy3 scored across 5 key signals, drawn from tasks submitted by a global community of users. Here’s an overview on the signals: User-satisfaction proxies - Confirmed Success: an explicit "yes that worked" feedback from the user - Praise vs. Complaint: implicit sentiment in users reactions - Steerability: can the model course-correct when you push back? Tool-use proxies - Bash Recovery: how it recovers from CLI errors (primary signal for tool use) - Tool Hallucination: does it call tools that don't exist Congrats to the @TencentHunyuan team on this release!

Kimi K3 is second only to Fable 5 on AA-Briefcase, our agentic knowledge work benchmark, but costs more than Opus 4.8 to run while averaging nearly an hour per task Last week @Kimi_Moonshot released Kimi K3, a 2.8T parameter model that scores 57 on the Artificial Analysis Intelligence Index, comparable to models such as Opus 4.8 and GPT-5.5. On AA-Briefcase, Kimi K3 scores an Elo of 1543, a +727 improvement over Kimi K2.6 and the second highest score recorded, behind only Claude Fable 5 (1574) AA-Briefcase is our new proprietary benchmark for agentic knowledge work, testing models on a fully private dataset of realistic tasks across thousands of complex input files. Tasks require deliverables such as spreadsheets, presentations, and UI mock-ups, with performance combined into a single AA-Briefcase Elo based on correctness, analytical quality, and presentation quality Key results for Kimi K3 on AA-Briefcase: ➤ Second only to Fable 5: Kimi K3 achieves an AA-Briefcase Elo of 1543, the second-highest score overall, ahead of GPT-5.6 Sol (max, 1501), Claude Sonnet 5 (max, 1388), and Claude Opus 4.8 (max, 1347). This is a +727 improvement over the previous-generation Kimi K2.6 (816) and puts Kimi K3 only behind Fable 5 ➤ Strong objective and analytical performance, with comparatively weaker presentation: Kimi K3 achieves a rubric pass rate of 51%, second only to Claude Fable 5 (56%) and ahead of Claude Sonnet 5 (max, 42.3%) and GPT-5.6 Sol (max, 41.8%). It also records an analytical quality Elo of 1754, comparable to Claude Fable 5 (1744). Presentation quality is comparatively weaker, with a Presentation Elo of 1471, below GPT-5.6 Sol (max, 1660) and Claude Opus 4.8 (max, 1492) ➤ ~10x increase in Cost per Task: Kimi K3 averages a cost of $10.57 per task, a ~10x increase from Kimi K2.6, placing it among the most expensive models to run on AA-Briefcase. This is driven by model token pricing, increased output tokens and relatively high turn use, averaging 83 turns per task, versus 67 for Claude Fable 5 and 50 for GPT-5.6 Sol (max). Kimi K3 is priced at $3/$15 per 1M input/output tokens, with a 90% discount for cached tokens ➤ Averages nearly an hour per task: Kimi K3 has an average Time per AA-Briefcase Task of 56.4 minutes. This is driven by a high number of turns, high output token use, and lower speeds using the first-party Kimi API

Kimi K3 on my shader test: "create a visually interesting shader that can run in twigl-dot-app make it like an infinite city of neo-gothic towers partially drowned in a stormy ocean with large waves." "Make it better" Very good model, not Sol Max or Fable, but great open weights

Laguna S 2.1 performs at GLM-5.2 level on building popular games with 6x fewer params! We gave three local models the same task: build three popular arcade games that play themselves. Each game is one self-contained HTML file with a bot that plays it. Prompts: – Geometry Dash – Doodle Jump – Air Hockey Outputs: Laguna S 2.1: 10.3K tokens GLM-5.2: 26.4K tokens Hy3: 10.4K tokens Laguna held its quality against a 753B model. We think Laguna's Geometry Dash looked the best of the three, the cube clears every spike and block. GLM won Air Hockey. Its table looked the most detailed of all. Hy3 was the only model that added shooting to its Doodle Jump. But Laguna is the only model in our benchmark that runs on a MacBook with 128GB!

I’ve gotten so used to Grok’s speech-to-text that typing now feels like hell Quick tip: Even when you’re using another agentic coding harness, try Grok STT through Grok Build Anthropic’s speech-to-text is still pretty basic actually in my experience. With longer technical prompts, it can quickly go off the rails and completely distort what you were trying to say Just press Ctrl + Space, start speaking, and let Grok transcribe the entire thought in real-time You can then copy it into another tool or press Enter and send it directly to Grok 4.5 right there in Grok Build It’s fast, highly accurate and paired with one of the most capable and efficient frontier models available Long, detailed rambling is often exactly what an AI agent needs to understand what you actually want