@atmorouter.dev: Tencent just shipped Hy4 flash, and the price per task is the headline 💸 Released Nov 16, Hy4 flash is Tencent's most cost-effective Hy model so far. What you get: • 224B total parameters (incl. 64B Engram), just 12B active per token • Natively multimodal, up to 1M tokens of context • $0.070 / 1M input, $0.281 / 1M output, $0.003 / 1M cached input (Hy4: $0.834 / $2.501) • 150+ tokens/s decoding (megakernels, 4-bit quantization, DFly) Tencent's blind test (85 internal WorkBuddy users, 259 tasks, 27 categories): • Hy4 flash: 2.79/4 satisfaction, ¥0.26 per task • Claude Haiku 5.5: 2.81/4, ¥4.73 per task • External models ran on their high tier The catch: on Tencent's internal WorkIndex, Hy4 flash scores 42.1 vs Hy4's 51.4, and Tencent still recommends Hy4 for long-horizon software engineering. All numbers are vendor-reported, so test on your own workload. Availability: Tencent Cloud TokenHub (50% off through Nov 30), free in WorkBuddy through Nov 30, OpenRouter "coming soon", open weights planned. Sources: https://hy.tencent.ai/research/hy4-flash https://www.tencent.com/tencent-launches-hy4-flash-to-power-high-speed-cost-efficient-ai-deployment-for-everyday-workflows/ Price per task is what you actually pay. AtmoRouter is the cheapest AI API gateway: one OpenAI-compatible endpoint, 100+ models, up to 95% off official price. atmorouter.dev #AI #LLM #Tencent #AIEngineering #AIAgents #LLMOps #GenAI #AIPricing

atmorouter.dev
atmorouter.dev
Open In TikTok:
Region: ID
Tuesday 17 November 2026 00:58:00 GMT
427
5
0
0

Music

Download

Comments

There are no more comments for this video.
To see more videos from user @atmorouter.dev, please go to the Tikwm homepage.

Other Videos


About