The Futurist

One-command LLM censorship removal — now with reproducibility

“Local AI sovereignty means having full control over model behavior — safety alignment included. As frontier model weights become widely available, tools like Heretic will be part of every serious local AI stack. The reproducibility features are a step toward professional-grade local inference.”

1.6T open-source MoE that nearly matches frontier — MIT, 1M token context

“The efficiency breakthrough is the story. If 1M-token context now costs 73% less to serve, that changes the economics of an entire class of applications. DeepSeek is compressing the frontier timeline faster than anyone predicted a year ago.”

Google's open multimodal models — vision, audio, and text under Apache 2.0

“The 100,000-variant Gemmaverse is a real ecosystem flywheel. Every new Gemma release compresses capability curves downward — things that required cloud APIs last year now run on-device. Gemma 4's audio addition makes it the first truly comprehensive local AI.”

27B dense coding model that outperforms models 10x its size on benchmarks

“The efficiency trajectory here is remarkable. A 27B model doing flagship-level coding work signals that the parameter-count ceiling for capable local models is lower than anyone expected two years ago. This democratizes AI-assisted development for individual developers and small teams who can't afford cloud API costs at scale.”

104B MoE model with only 7.4B active params — big model quality at small model speed

“The proliferation of high-quality, truly free open-weight models is one of the most significant structural shifts in AI right now. Ling-2.6-Flash represents Chinese AI labs maturing to the point of producing globally competitive open releases — which accelerates the entire ecosystem and drives down the cost of intelligence for everyone.”

1.58-bit LLMs that run at 82 tok/s on M4 Pro and on your iPhone

“On-device AI at 27 tokens per second on a phone is the inflection point that makes LLMs a platform primitive rather than a cloud service. Once inference is this cheap and fast on commodity hardware, the entire economic model of AI-as-API-call collapses. Ternary quantization is an early signal of where efficiency research is heading.”

35B total, 3B active: Alibaba's lean MoE coding beast goes fully open source

“The gap between open and closed models is closing faster than anyone predicted. When a freely downloadable model matches Claude Sonnet on multimodal benchmarks, the frontier lab pricing power evaporates. Qwen3.6-35B-A3B is another milestone in the commoditization of intelligence — and commoditization always accelerates adoption.”

1.58-bit LLMs that fit in 1.75 GB — runs in your browser via WebGPU

“Browser-native LLMs with no server change the entire privacy calculus. If this scales to 13B+ parameter territory at comparable compression ratios, every personal AI assistant can run offline on consumer hardware. That's a trajectory worth tracking closely.”

First commercially licensed 1-bit LLMs — 8B in 1.15 GB, 8x faster on-device

“Billions of devices cannot run even 4-bit quantized models. Bonsai makes LLM inference feasible for the embedded world — the next billion AI interactions won't happen in the cloud. If PrismML's quality curve improves with larger models, this is the beginning of the post-cloud LLM era for edge computing.”

3B-parameter open model supporting 70+ languages — runs offline on a phone

“The 5 billion people who don't speak English as a first language are the next wave of AI users — and they'll largely be on mobile, offline-capable devices. Tiny Aya is building the infrastructure for that wave. The region-specific model design suggests Cohere Labs is thinking seriously about this rather than treating multilingual support as a checkbox.”

1-bit quantized 8B LLM — 1.15GB, runs on-device at 368 tok/s

“1-bit LLMs running on-device are the foundation for truly private, always-available AI. When an 8B model fits in 1GB and runs on a phone, every app becomes AI-capable without cloud dependencies. Bonsai-8B is a milestone in the long march toward AI that runs everywhere.”

399B open MoE reasoning model that's 96% cheaper than Claude Opus

“A US-built, Apache-licensed frontier reasoning model competitive with closed offerings fundamentally changes the open-source AI landscape. The talent and capital required to do this was thought to only exist at the biggest labs. Arcee just proved otherwise.”

Google's first Apache 2.0 open model family with native multimodal

“Native multimodal understanding — including audio — on models small enough for phones changes what ambient computing looks like. Gemma 4 on-device could be the model layer for a generation of always-on smart devices that don't need cloud inference.”