--- id: "2026-11-16-tencent-hy4-flash" url: "https://postcutoff.com/e/2026-11-16-tencent-hy4-flash/" as_of: "2026-11-22T13:32:00+01:00" date: "2026-11-16" date_precision: day category: model-release importance: 2 confidence: high status: [Confirmed] sources: 5 editor: human editor human_review: null version: "2026-11-19" --- As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-11-16-tencent-hy4-flash/ # Tencent Hy4 flash Full title: Tencent releases Hy4 flash: 224B-parameter (12B active) MoE with 1M context at $0.07/$0.28 per million tokens; weights promised Tencent released Hy4 flash on Nov 16, 2026, the small sibling of its Hy4 model, aimed at everyday office work: 224B total and 12B active parameters, 1M-token context and multimodal input, priced at $0.07 input / $0.281 output per million tokens (about 2 yuan per million output). Tencent says it cost one-twentieth as much per task as full Hy4 on its WorkIndex benchmark and plans to open the weights. ## Key facts - Architecture (TechNode, Pandaily): 224B total parameters, 12B active per forward pass; up to 1M tokens of context; multimodal inputs - API price: $0.07 per 1M input tokens, $0.281 output, $0.003 cached input; half price on Tencent Cloud TokenHub at launch; free in WorkBuddy until Nov 30; also in Yuanbao - Tencent's tests: on its internal WorkIndex, cost per task 1/20 of full Hy4 (the '95%' headline); in a blind test, 85 WorkBuddy users rated 259 tasks and Tencent says it came close to Claude Haiku 5.5 on satisfaction with the lowest cost and shortest completion time - Open weights and OpenRouter listing promised, no date (TechNode) - Same day Tencent published notes from an anniversary all-hands urging core businesses to rebuild around AI and fewer management layers, naming WorkBuddy 'the company's first AI-native product to gain traction' (TechNode via IT Home); no headcount figures - Weights (Nov 17): tencent/Hy4-flash, Hy4-flash-base and Hy4-flash-FP8 on Hugging Face (~220B parameters per safetensors metadata, licence 'other'; 281 likes by Nov 19); TechNode: Tencent says Hy4 flash passed 50 trillion tokens of first-day usage, and Tencent is cutting Yuanbao app spending after it fell short of expectations ## What happened Tencent had announced Hy4 and Hy4 flash at its Oct 29 ecosystem conference; full Hy4 came out with open weights on Nov 4, and the flash model followed on Nov 16 as a cheap model for research, slides and spreadsheets inside Tencent's WorkBuddy agent. ## Why it matters It is among the cheapest 1M-context models from a major lab, and part of a price war among Chinese labs for agent workloads. ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 169 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 139 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 230 days after its cutoff - Grok 4.7 (training cutoff May 2026): 169 days after its cutoff ## Sources 1. [Hugging Face: tencent/Hy4-flash](https://huggingface.co/tencent/Hy4-flash) (huggingface.co, code) 2. [TechNode: Tencent releases Hy4 flash model weights for local deployment](https://technode.com/2026/11/19/tencent-releases-hy4-flash-model-weights-for-local-deployment/) (technode.com, press) 3. [TechNode: Tencent launches Hy4 flash, cutting AI task costs by 95% in internal tests (Nov 16)](https://technode.com/2026/11/16/tencent-launches-hy4-flash-cutting-ai-task-costs-by-95-in-internal-tests/) (technode.com, press) 4. [Pandaily: Tencent releases Hy4 flash, a 224B model at 2 yuan per million output tokens](https://pandaily.com/tencent-hunyuan-hy4-flash-224b-12b-active-1m-context-workindex) (pandaily.com, press) 5. [TechNode: Tencent publishes all-hands notes urging teams to rebuild their businesses around AI (Nov 16)](https://technode.com/2026/11/16/tencent-publishes-all-hands-notes-urging-teams-to-rebuild-their-businesses-around-ai/) (technode.com, press) ## Changes - 2026-11-19: Weights released on Hugging Face - 2026-11-16 (filed): Created (news agent) ## Related - 2026-11-04: [Tencent releases Hy4 as an open-weight model with vision and computer use, claiming top-tier status among Chinese open models](https://postcutoff.com/e/2026-11-04-tencent-hy4-open-weights/index.md) - 2026-08-28: [Tencent open-sources Hunyuan Hy4 preview (770B MoE, 1M+ context)](https://postcutoff.com/e/2026-08-28-tencent-hunyuan-hy4-preview/index.md)