Loading…

🎙 Tencent's 7 cent model is real, the 50 trillion is its own count | DATASPHERES AI - Dataspheres AI

Tencent's 7 cent model is real, the 50 trillion is its own count (2 min), with captions The story Tencent released Hy4 flash on November 16, 2026, and two ...

Tencent's 7 cent model is real, the 50 trillion is its own count (2 min), with captions The story Tencent released Hy4 flash on November 16, 2026, and two days later said the model had handled more than 50 trillion tokens on its first day, as TechNode reported on November 19, 2026 . That figure is Tencent's own count. No source read for this edition reports an independent check of it. A separate public count exists for one channel: OpenRouter's rankings (read November 19, 2026, usage through November 18) place Hy4 flash first for the most recent complete day, at 23.8 trillion tokens routed through OpenRouter. That is a different day and a single channel, so it neither confirms nor contradicts Tencent's number. The price is on the price list. Hy4 flash lists at $0.07 per million input tokens and $0.28 per million output tokens, per Dataconomy (November 17, 2026) , which also reports a 50 percent discount and free access in Tencent's WorkBuddy app until November 30. GPT-6.2 Sol lists at $2 and $10, per BetaNews (November 19, 2026) and OpenAI's pricing page. Gemini 4 Neon lists at $3 and $15, in Google's announcement of November 18, 2026 . Tencent published the model's weights on November 18, 2026, on Hugging Face . Open weights mean anyone can download the model, run it on their own machines and fine-tune it, and that other providers can host it. Two limits apply. The weights come under the Tencent Hy License, which adds commercial conditions for large businesses that sell model access, as Digital Applied describes (November 19, 2026) . And the smaller published version of the weights takes 287 GB, according to Neurohive (November 18, 2026) . On quality, the independent number and Tencent's number differ. Vals AI (read November 19, 2026) ranks Hy4 flash 36th of 42 models on its Vals Index, at 50.8 percent, against 66.1 percent for GPT-6.2 Sol and 70.9 percent for Gemini 4 Neon. Running that test cost $0.07 on Hy4 flash, $3.18 on Sol and $7.40 on Neon. Tencent's claim that Hy4 flash performs about as well as a rival model comes from its own blind evaluation with 85 internal users, described on its model card. Tencent also says the model writes more than 150 tokens per second; Neurohive reports about 70 measured on OpenRouter. The hosts' own call, which is a prediction and not a reported fact: with the weights open, another provider lists Hy4 flash below 7 cents by spring. Highlights Highlight Listen The episode as audio