AI Models & Tools

vLLM’s Kimi K3 decode path reaches 842 tokens a second in a specific GB300 test

vLLM reports 842 versus 488 aggregate decode tokens a second for Kimi K3 on 16 GB300 GPUs under a specific DSpark setup. Prefill is excluded.

UfiSpace shows liquid-cooled 1.6T switch at OCP summit

UfiSpace demonstrated an early-access 102.4 Tbps switch with liquid cooling at OCP. The public material establishes a live demo, not a deployed or independently measured data-center system.

Replit names OpenAI its preferred Agent partner while Claude remains available

Replit CEO Amjad Masad says OpenAI is now the preferred partner and default for Agent, while Claude remains an option. Replit's model controls show how that choice varies by mode and plan.

FLUX 3 Image Klein: open weights, API access and license limits

Black Forest Labs has released FLUX 3 Image Klein weights and a hosted API. Its model package, runtime memory and commercial permissions differ in ways teams need to check.

RTX Spark laptops reach launch, with software on separate schedules

NVIDIA says RTX Spark laptops became available October 16. Its launch pages disagree on starting price, while memory options and AI software stages vary by product.

Continues Microsoft's Windows AI launch spans available tools and future Copilot features

Run distributed MLX inference across Mac Studios

Apple's MLX and Thunderbolt RDMA setup can split model inference across Macs. This documentation-based walkthrough shows the prerequisites, benchmark path and limits of Apple's reported speeds.

StudentSim tests AI tutor feedback without measuring student learning

StudentSim compares AI student simulators with learner records and tests responses to guided corrections. Its chess tutor study rates advice, not learning gains in real students.

Audit one coding-agent session with AgentsView

Find a known local coding-agent transcript in AgentsView, inspect its logged tokens and cost provenance, and read a bounded day report without treating estimates as bills.

Claude Haiku 4.5 leaves consumer model selectors on October 29

Anthropic will remove Haiku 4.5 from Free, Pro and Max model selectors on October 29. Existing chats remain, but resumed chats will answer with Haiku 5.5.

Continues Claude Haiku 5.5 cuts token prices but changes the migration math

Claude proposes quantum circuits that reduce estimated gate and qubit costs

An Anthropic and Harvard preprint uses Claude to propose quantum circuits and a separate evaluator to check them. Seven of eight tasks show lower estimated gate and qubit counts under the paper's comparisons.

How BlenderWorld evaluates editable 3D scenes made by AI agents

QwenLM's BlenderWorld benchmark asks agents to submit native Blender projects. Its task rubrics and visual judges test different aspects of a 3D scene, with limits for real production use.

Junie Local gives Windows developers a managed local coding agent

JetBrains has made Junie Local stable on supported Windows RTX hardware. Its CLI connects to a local model server, while developers still control repository access, commands and review.

Review a contract against a playbook with Copilot in Word

Microsoft's legal skill in Word reviews a contract against a .docx playbook, shows extracted rules and proposes tracked edits. A lawyer must check the findings and choose each change.

Use paired USB traces to investigate an unsupported Linux device

A documentation-based guide to recording the same hardware operation under a Windows vendor driver and a Linux driver, then testing the differences on real hardware.

How OpenRouter Armor screens tool calls, and when the client must stop them

OpenRouter Armor classifies proposed tool calls against developer rules. Responses and Chat Completions can withhold flagged calls; Messages compatibility leaves the client responsible.

Check Android chat scam alerts and Google Messages spam protection

Check the current Android chat scam warning and Google Messages spam settings, with eligibility, privacy details and the October 2026 features still rolling out.

Microsoft’s MAI-Base-2 report measures pretraining, not a released model

Microsoft reports MAI-Base-2's held-out text results and fitted training-efficiency estimates. The base model is still in post-training, with no public checkpoint to test.

Hy Memory 2.0 gives agents three ways to keep context

Tencent's Hy Memory 2.0.1 SDK separates chat, work and coding recall. Its MemHub app has source but no public release artifacts; integrations still need model endpoints and explicit access boundaries.