freshweights.com · Tue 17 Nov

Tuesday 17 November

Tencent releases Hy4 flash with 12 billion active parameters and lower task costs

Tencent headquarters towers under construction beside a prominent bus stop sign in Shenzhen

AI-generated summary Tencent released Hy4 flash on 16 November 2026 with 12 billion active parameters and a 1-million-token context window for routine workplace tasks. Tencent says the model is available on platforms including Yuanbao and WorkBuddy, with free access for WorkBuddy continuing through 30 November.

Also covered byFresh WeightsRSS.bze/acc
Earlier:

DAILY AI BRIEF 🗞 — Nov 17 MODEL FORECAST 👀: - MAI-Code-2 and MAI-Code-2-Flash, today!

AI-generated summary MAI-Code-2 and MAI-Code-2-Flash appeared in GitHub Copilot's model catalog with 1.05M context according to reports. Microsoft updated MAI Playground with a redesign ahead of the Ignite keynote where new models are expected.

Also covered byUnite.AIRR Tech PulseFresh Weights

Sarvam Token Factory: India-Hosted AI Inference With an OpenAI-Compatible API and Rupee Pricing

AI-generated summary Bengaluru-based Sarvam AI has introduced Sarvam Token Factory, a paid inference service that runs its own models and popular open-weight models on GPUs in India using an OpenAI-compatible API. Sarvam states that its service charges in rupees per million tokens, supports zero data retention on request, and hosts models such as Sarvam 105B, GLM 5.3, DeepSeek V4.1 Flash, and Gemma 4 31B.

Also covered byMarkTechPostThe News92Fresh Weights

Liquid AI d1.2 adds audio and typed answers to its API

AI-generated summary Liquid AI released d1.2 and d1.2-ultra on November 16, adding audio and structured answer types to its hosted decision models available through its API. Liquid reports that d1.2-ultra scores 75 on Artificial Analysis’s Decision Index v2.1 and states that internal evaluations using held-out public datasets cover the five new task types.

Also covered byAI PingFresh Weights