@rohanpaul_ai: More good news for local LLMs. Tencent’s new Hunyuan Hy3 reaches Gemini 3.5-level physics quality for 35x less cost. Te…
Summary
Tencent's new Hunyuan Hy3 model achieves Gemini 3.5-level physics quality at 35x lower cost, based on a test comparing simulations in atomic.chat.
View Cached Full Text
Cached at: 07/06/26, 08:19 PM
More good news for local LLMs.
Tencent’s new Hunyuan Hy3 reaches Gemini 3.5-level physics quality for 35x less cost.
Test was done on atomic[.]chat, a desktop app that runs LLMs locally.
The prompt asked 4 models to build bowling, air hockey, and pool simulations. The harder part was preserving physical cause and effect.
A strike needs collision timing, mass transfer, pin rotation, friction, and believable scattering.
A pool break exposes the same weakness, because every wrong angle compounds immediately.
Interestingly, DeepSeek-V4 spent the highest number of tokens (50,600 ), yet produced the weakest visual physics in this test.
atomic.chat (@atomic_chat_hq): New Hunyuan Hy3 hits Gemini 3.5 quality on physics for 35x cheaper!
We gave 4 models the same prompt: build three self-contained HTML5 canvas scenes with real physics demos
Prompts:
- A bowling ball knocking down the pins
- An air hockey rally that ends in a goal
- A pool
Similar Articles
@TencentHunyuan: Hy3 is here. 295B MoE. Best in its size class. Rivals trillion-scale flagships. Reliable and affordable for most agenti…
Tencent Hunyuan releases Hy3, a 295B MoE model that rivals trillion-scale flagships, open-sourced under Apache 2.0 with a free API for two weeks.
@rohanpaul_ai: Another good news for local-LLM from atomic[.]chat, that runs 100% offline on your computer. They just showed MTP (Mult…
atomic.chat's MTP technique speeds up local LLM inference by drafting multiple tokens and verifying them together, achieving up to 137% speedup on Qwen 27B dense model with zero accuracy loss.
@atomic_chat_hq: 1-bit Hy3 running locally is 2.2x faster than its API at the same quality! We gave both models the same task and compar…
Tencent's Hy3 295B model now available in 1-bit and 4-bit GGUF formats, achieving 2.2x faster local inference compared to cloud API while maintaining quality, as demonstrated by running on 4x RTX 5090 with 128GB VRAM.
@ShunyuYao12: Hy2 -> Hy3 preview -> Hy3 Another massive leap forward, under half a year. Not just a leap of reasoning or agentic capa…
Tencent releases Hy3, a 295B MoE AI model that rivals trillion-scale flagships, open-sourced under Apache 2.0 with a free API trial.
@cline: Tencent Hy3 represents a breakthrough in open source intelligence. It's topping benchmarks at only 295b params / 21b ac…
Tencent Hy3 is a new open-source frontier-level AI model with 295 billion parameters (21 billion active) that tops benchmarks and can run locally on consumer hardware. It is available for free in Cline.