freym
freym / blog

Model news, explained

Articles on new AI model releases and what they mean — written from the official announcements we track in the live news feed and the model catalog.

Updated 2026-08-08

Seedance 2.5: the model every AI video platform raced to ship

Within ten days, Runway, Pika, Higgsfield, Leonardo, Hedra and fal.ai all announced Seedance 2.5. Here is the full adoption timeline — and what it says about the AI video market.

Qwen 3.8-Max: Alibaba's 2.4-trillion-parameter push to the frontier

Qwen 3.8 launched at 2.4T parameters with open weights promised. Three weeks later, Qwen 3.8-Max ranks #1 on the Agentic Index and #5 on the AI Index. The full timeline.

The open-weights wave: why every lab suddenly gives models away

In three weeks, MiniMax, Higgsfield, Alibaba, and Tencent all open-sourced serious models — video models that run on an RTX 5090, a 2.4T-parameter flagship, translation models topping Hugging Face. What changed.

GPT-5.6 Sol, Luna and Terra: OpenAI's August reshuffle, explained

An 80% price cut for Luna, 68% fewer factual errors for Sol, extended thinking free for everyone, and a reasoning-effort slider. What OpenAI changed in the GPT-5.6 family and who gets what.

FLUX 3 and FLUX 3 Video: Black Forest Labs goes multimodal

BFL's FLUX 3 unifies image, video, audio and action-prediction in one architecture. FLUX 3 Video generates picture and sound in a single pass, up to 20 seconds. The full release timeline.