Google's new workhorse model launches at $0.75 per million input tokens and $3.75 per million output tokens through December 31. The listed January price is twice the launch tariff, while the performance gains remain Google claims. Synthetic narration produced by Tachibana Navi.

DeepSeek has filled in its earlier price-rise warning with higher V4-Flash and V4-Pro tariffs plus two daily peak windows. Delay-tolerant jobs now have a cheaper time to run. Synthetic narration produced by Tachibana Navi.

Sakana Chat now lets users try Fugu’s dynamically coordinated models alongside an updated Japanese-focused Namazu, sandboxed Python and new file tools. Synthetic narration produced by Tachibana Navi.

Mercari’s company-wide Claude rollout uses role tiers, temporary administrator access and a central model gateway. The interesting security decision is not simply what the agents are forbidden to do, but how easily staff can use an approved route. Synthetic narration produced by Tachibana Navi.

Google will sell Pixel 11 in Japan from August 20, starting at ¥136,800. Gemini Intelligence, which Google says the phone is designed for, has no Japanese arrival date. Synthetic narration produced by Tachibana Navi.

Manus says it will return to independent operation as its separation from Meta is completed. For some users, the corporate unwind has an immediate cost: back up recent data, lose access for two days, then restore it. Synthetic narration produced by Tachibana Navi.

Square Enix routed, filtered and retried its Gemini-powered companion—then gave it a fixed apology for times when retries still failed to produce a usable response. Synthetic narration produced by Tachibana Navi.

Muse Glimmer is an Apache-2.0, 29.6B-parameter agent model whose 16.76 GB quantized weights ship beside separate vision and speculative-decoding components; Meta targets the complete configuration at 24 GB VRAM. Synthetic narration produced by Tachibana Navi.

A preregistered study found that parents’ willingness to pay for unrestricted AI rose by more than 60 percent as hypothetical peer adoption increased, even when risk information strengthened support for a collective ban. Synthetic narration produced by Tachibana Navi.

FineBooks compares 14 open OCR models on expert-transcribed historical pages, exposing the different standards required for training corpora, scholarship and library systems. Synthetic narration produced by Tachibana Navi.