Frontier Models Clash
Kimi K3 shakes frontier. Moonshot AI’s open-weight Kimi K3 competes with top closed models, beating Claude Fable 5 on some benchmarks and reigniting the open vs. closed debate.
- → Kimi's open model K3 nears GPT-5.6 Sol and Fable 5 while signaling the end of super cheap Chinese AI
- → Kimi K3, and what we can still learn from the pelican benchmark
- → [AINews] Kimi K3 2.8T-A50B: the largest open model ever released; Opus 4.8-class at Sonnet 5 pricing
- → Kimi K3 weights to be released on the 27th.
- → Kimi K3 ranks #1 on @AfterQuery's SpreadsheetBench 2, surpassing Claude Fable 5
- → Kimi: Threat or menace?
- → kimi.ai teasing a video with lots of 3's in it
- → [AINews] not much happened today
- → Kimi moment. I think the writing is on the wall for Anthropic and OpenAi
- → Kimi K3 is currently at the top of the leaderboard for Text Arena filtered for science queries.
- → Kimi K3 (max) beats Sonnet 5 on Simple Bench
- → Kimi K3 is top of nextjs eval
- → Just like Deepseek, China's Kimi K3 is forcing Western AI labs to question their compute advantage
- → Kimi K3 🌕, Gemini 3.5 delayed ⏳, crushing ARC-AGI 3 🤖
GPT-5.6: Power and Peril. The GPT-5.6 family brings programmatic tool calling and parallel subagents, but its tendency to unwittingly delete files and databases underscores the need for sandboxing.
- → The Sequence Radar #893: Last Week in AI: GPT-5.6, Grok 4.5, Muse Spark 1.1 and the Post-Chatbot Stack
- → OpenAI’s new flagship model deletes files on its own, people keep warning
- → GPT-5.6 is deleting user files when given full access, and OpenAI says it shouldn't but did
- → [AINews] Codex usage up >10x in 6 months to 7M users, +1M in the past ~day; did Codex overtake Claude Code??
Claude Fable 5 extended. Under pressure from GPT-5.6 and Kimi K3, Anthropic reversed plans and kept Fable 5 on paid plans, directly competing on accessibility.
DeepSeek V4 looms. DeepSeek is teasing V4 with low-cost API and open weights, while community optimizations already let its flash variant run on consumer GPUs at usable speeds.
Agents Enter Real-World Workflows
Browsing agents arrive. Anthropic gave Claude Code a built-in web browser with safety classifiers, and Cursor launched a general agent, moving coding assistants toward autonomous task execution.
Agents handle real transactions. DoorDash’s agent architecture boosted conversions by 24%, and Stripe’s benchmark shows agents can code integrations but often fail at validation, highlighting reliability gaps.
Production agents stumble. Most enterprise agents remain chatbots; 54% of firms had agent security incidents, and experts argue agents need cloud-native operational primitives like microservices.
- → How to Debug Coding Agents with LangSmith Traces
- → The agent security gap: 54% of enterprises have already had an AI agent incident, and most still let agents share credentials
- → The AI context gap: Enterprise AI organizations have a trust problem, not a retrieval problem — and most are still building the fix
- → The agent evaluation gap: Enterprise AI organizations have a reality-alignment problem, not a coverage problem — and most are shipping to production anyway
- → Cloud Native Infrastructure Emerges as the Foundation for Trustworthy Agentic AI
- → QCon AI Boston: Production AI Moves Beyond Prompts to Platforms, Harnesses, and Evals
AI Infrastructure, Funding, and the Data Center Race
Data centers hit resistance. S&P downgraded Oracle over AI capex risk, while local protests and New York’s moratorium on hyperscale data centers reflect growing pushback against AI infrastructure.
AI funding surges. Databricks raised at a $188B valuation, Meta is negotiating a $10B data center lease to Anthropic, and Nous Research secured $75M for its open-source agent.
Energy and hardware bets. Energy firms raised the most since the dot-com era to power AI, and a $400M loan backed by SambaNova chips signals growing investment in non-GPU AI hardware.
On-Device AI Breakthroughs
1-bit models hit phones. PrismML’s Bonsai compresses Qwen3.6-27B to 3.9GB while retaining 90% of benchmark scores, enabling on-device tool-calling agents; Apple is reportedly in talks.
- → Bonsai 27B: The First 27B-Class Model to Run on a Phone
- → Bonsai 27B: 1-bit dense LLM running locally in your browser using custom WebGPU kernels
- → Prism-ML Bonsai Qwen 3.6 27B
- → PrismML’s new Ternary Qwen3.6 27B runs near fp16 precision on 10GB of memory!!!
- → So what's the consensus on 1bit models? Is it still a pipe dream?
- → PrismML Bonsai 27B is surprisingly usable on the Jetson Orin Nano 8GB
- → Is anyone having any luck with the Ternary Bonsai 27B DFlash?
- → Can we get a "not base model" flair?
- → Bonsai 27B runs locally on an iPhone - a 27B model in 3.9GB
- → User experience of Bonsai-Ternary-27B on 4060Ti 16GB for KB management and productivity assistant use cases
- → Bonsai 27B is a full open reasoning model that fits on an iPhone
- → Bonsai-27B & Ternary-Bonsai-27B - Updates (on PRs)
- → Apple in talks with startup PrismML that shrinks AI models to run on an iPhone
Home labs run frontiers. llama.cpp optimizations like DFlash accelerate MoE models up to 6x, and a OnePlus phone streamed experts from flash to run a 60GB model at 1.3 tok/s.
- → I tested all llama.cpp's speculative decoding methods on Qwen 3.6 27B: MTP ~2.7x, DFlash ~3.7x, n-gram stack ~6x on real coding. Local AI win. My findings on RTX 6000 PRO.
- → DFlash makes Qwen3.6 27B 2.2x faster with no quality loss
- → GPT-OSS-120B, Qwen 30B and Gemma 26B on an Android phone at 1-5 tok/s: +60GB model, 11GB of RAM, CPU only
Policy, Lawsuits, and the Open-Source Standoff
Apple–OpenAI legal fight. Apple sued OpenAI for trade secrets, alleging poaching of over 400 employees, including the chief hardware officer, threatening OpenAI’s IPO and cloud trust.
- → Apple sues OpenAI after ex-engineer allegedly used bug to steal trade secrets
- → The wildest allegations in Apple’s trade secrets lawsuit against OpenAI
- → The 6 wildest claims in Apple’s lawsuit against OpenAI
- → OpenAI pushes back on Apple trade secret lawsuit
- → Sam Altman didn’t need another lawsuit
- → How Apple’s big lawsuit could disrupt OpenAI’s IPO plans
- → Apple’s plot to crush OpenAI
- → Apple’s lawsuit couldn’t come at a worse time for OpenAI
Chinese open-source surge. Chinese models now account for 41% of HuggingFace downloads, Kimi K3 rivals Western leaders, and China launched a parallel AI governance body excluding the West.
- → The real AI race may no longer be at the frontier
- → Source: the Trump administration and industry groups discussed streamlining US open model releases of equal or lesser capability to leading Chinese open models
- → Chinese President Xi Jinping speaks at World AI Conference and reaffirms commitment to open source to promote"openness and win-win"
- → China’s Xi Touts Open-Source AI and Takes a Swipe at U.S. Dominance
- → China's new World Artificial Intelligence Cooperation Organization is President Xi's clearest play yet for a parallel AI order
Accountability demands grow. German regulators held chatbots liable for content, xAI sued a user for generating CSAM, and Demis Hassabis proposed a FINRA-like body for frontier safety reviews.
That's the week in review - see you next Sunday.