Skip to main content

Posts

Showing posts from October, 2026

Liquid LFM3: A Small Model Needs a Clear Output Contract

Image
The first JSON object looked usable until I passed it to a parser. The order number and quantity were there. A closing code fence after the object was enough to break automatic ingestion. Liquid LFM3-1.7B is an interesting small local model, but I would give it a narrow job and an explicit output contract. In twelve synthetic extraction and routing checks, four unconstrained responses passed my strict check. With the same prompts and a JSON Schema, all twelve passed. That is a small application exercise, not an estimate of general model accuracy. What did Liquid AI release? Liquid AI announced LFM3 on October 27, 2026. It is a family of roughly 2B-parameter models branching from a shared backbone, with separate checkpoints for different inputs. The 1.73B text model is the one I ran here. Downloading it does not give that checkpoint image, video or speech input. Release announcement · Family guide . Image: Liquid AI, from its official LFM3 model card. This is a brand banner, not a mo...

Mac Studio 512GB: Can It Run a 1T-Parameter AI Model?

Image
A trillion-parameter coding model on one Mac was the first thing I wanted to check. Kimi K2.7 Code makes that a plausible project: one published Q3 package, including its vision projector, is about 419.8GB . The catch is in the word “package.” That number leaves out the operating system, working buffers and conversation cache. The 512GB Mac Studio opens a serious door for giant local AI. I narrowed this comparison to five models with roughly 480 billion to one trillion parameters, rather than filling it with smaller models that miss the point. I checked downloadable weights, exact file sizes and documented runtime routes. This is a capacity and compatibility analysis; I have not benchmarked these models on a 512GB Mac. Which 512GB Mac Studio is actually available? Apple's current option is the M5 Ultra with a 36-core CPU and 80-core GPU . Its listed memory bandwidth is 1.2TB/s. The generation was announced on August 25, 2026, with general availability from September 22; Apple s...

Gemini 4 Carbon: 59 Points and a $2/$10 Intro Price

Image
The line I keep coming back to in Google's Carbon announcement is the pricing footnote: the introductory rates will eventually double. A model can look attractive on launch day and produce a very different budget later. That matters especially when the pitch is work that keeps running for hours or days. Google announced Gemini 4 Carbon on October 28, 2026 . Access begins with selected cyber defenders; a general consumer release date is not given. It also tops an independent intelligence index. Those are useful facts, but they answer different questions: capability, availability and the cost of actually finishing a job. Google’s official Carbon announcement artwork. Image: Google. This is promotional art, not a generated model output. Who can use Gemini 4 Carbon right now? Google is expanding access through its Fairwind Program . DeepMind describes that program as early access for trusted, high-priority defenders, including governments, healthcare providers and telecommunications s...

GPT-6.1 Terra: What It Does, Who Gets It, What It Costs

Image
The detail I kept coming back to in the Terra launch was where you can use it. GPT-6.1 Terra arrived on October 26, but looking for it in ordinary Chat will get you nowhere. It is available in Work and Codex. That distinction tells you a lot about the job OpenAI wants it to do. The official rollout note places images, screens and actions together. My interest is practical: can it read something on screen, make the right change, and leave a result I can verify? The release makes Terra worth trying for that loop. It does not, by itself, establish how reliably Terra completes your own work. This is a source-based launch analysis, with my own cost calculations and a proposed test protocol, rather than a hands-on review. OpenAI’s official Terra artwork. Image: OpenAI documentation; branding, not a model-generated sample or performance result. Where can you actually select GPT-6.1 Terra? OpenAI lists Plus, Pro, Business, Enterprise and Edu for the Work/Codex rollout. In the desktop app...

Kimi K3.1 Pro: A 3D File Still Has to Fit

Image
AGENTS & WORKFLOWS Open Source Factory · Oct 23, 2026 · 7 min read A downloadable 3D model gives you something to inspect, alter and print. It does not tell you whether two holes are the right distance apart. Kimi K3.1 Pro's 3D Make is most interesting at that boundary between a convincing shape and a usable part. In this article What would make an AI-generated 3D model useful? What does the 92.7 BenchCAD result actually measure? How can a shape look right and still be 0.5 mm wrong? What would I check before the first print? Which version can I use today? Measure the exported part Kimi K3.1 Pro is available through Kimi’s services, while full weights are promised by November 4. Its 3D Make feature offers online previews, editing and downloadable models. That is enough to make the release worth investigating. It does not establish that a replacement bracket will fit. Moonshot’s launch post supplies the product claims; the dimensional example below is my own calculation. I have...

1Password for Agents: Signing In Is Only the First Step

Image
POLICY & RESEARCH Open Source Factory · Oct 23, 2026 · 7 min read An agent needs to sign in before it can do useful work inside an account. Handing over the password creates an obvious problem; keeping the password hidden addresses only part of it. 1Password's proposal deserves a look at both the sign-in request and the access that remains afterward. In this article How can an agent sign in without seeing the password? Which agents can actually use it? Is it included in my current 1Password plan? What changes after the login succeeds? What would I check before connecting an account? Inspect access after sign-in That is what caught my eye in 1Password for Agents , announced on October 22. According to the company, a supported agent can ask to use a saved login, while 1Password handles the credential separately. The agent gets into the account without receiving the password. I find that useful. I would still want to know exactly what happens after the door opens. The launch ann...