Lemonade post-trained Hy3, a 295B model, to generate Roblox UI in its agentic Roblox Studio harness, targeting parity with Grok 4.5 at lower cost for its hundreds of thousands of users. The team built…
RAZOR, a training-free expert pruning method for mixture-of-experts (MoE) large language models, achieved the highest nine-task macro average among evaluated pruning methods in all eight settings test…
A developer added a WebMCP ON/OFF ablation benchmark to DeepDeck, an MIT-licensed macOS desktop project built on DeepSeek Harness, to measure whether exposing structured browser tools actually reduces…
Tencent AI Studio, the company's free consumer web studio for its Hunyuan (Hy) model family, has been added to the Uprouter directory as an unofficial web-cookie integration using session cookies from…
Tencent released Hy4 Preview, a large language model with a significant size increase from its predecessor Hy3, which had 295B total parameters, 21B active parameters, 256,000 context length, and 598G…
Tencent has unveiled Hy4 Preview, an open-source Mixture-of-Experts (MoE) model, marking a significant escalation in China's AI arms race. The model, which builds on the previous Hy3's 295 billion tot…
A developer's benchmark tests show that the cheapest model on their plan, MiMo-V2.5, loses all four benchmarks to a more expensive model but still beats mid-tier models that charge up to 14 times more…
OpenCode Go, a new low-cost subscription service from OpenCode, offers 8× usage limits for Hy3 for a limited time, priced at $5 for the first month and $10/month thereafter. The service provides relia…
Tencent's Hy3 scored 34.8% on the HTB-Challenger Benchmark, the third-lowest among all tested models, and got stuck on 9 of 16 Hack The Box challenges, generating a median of 101,509 output tokens per…
Tencent Holdings reported second-quarter revenue of 204.8 billion yuan, up 11% year-over-year, but net profit of 56.02 billion yuan missed analyst estimates of 58.51 billion yuan as capital expenditur…
Tencent Holdings reported an 11% rise in second-quarter revenue to 204.8 billion yuan ($30.36 billion), driven by AI-enhanced advertising and steady gaming income, but net profit rose only 0.7% to 56 …
Tencent has integrated its Hy3 large language model into the international edition of WorkBuddy, offering free access until Aug. 31, 2026, Pacific Time. Hy3 has 295 billion total parameters, 21 billio…
Tencent Cloud opened global access to its Hy3 reasoning model on August 5, integrating it into WorkBuddy, Tencent Design Miora, and Tencent Cloud TokenHub, alongside API access and downloadable weight…
Tencent opened its Hy3 large language model to global users on August 5, 2026, making it available through WorkBuddy, Miora, and Tencent Cloud TokenHub, with free access until August 31, 2026. Hy3, a …
Tencent Holdings rolled out its flagship artificial intelligence model Hy3 to global developers, enterprises, and users through WorkBuddy, Tencent Design Miora, and TokenHub, with free access on WorkB…
Moonshot's Kimi K3, a 2.8 trillion parameter open-weight model with weights promised for July 27, leads the July 2026 AI Model Radar, which also includes Thinking Machines Inkling (975B), Tencent Hy3 …
A developer highlights that Tencent's Hy3 model, a nearly 300 billion parameter AI, is free for the next five days on OpenRouter, urging users to integrate it with tools like Claude Code.…
Tencent Holdings released Hy3, its first full model under Chief AI Scientist Yao Shunyu, on July 6, activating 21 billion of 295 billion parameters per call and cutting hallucination rates from 12.5% …
Tencent released Hy3, a 295-billion-parameter open-weights AI model, at roughly six cents per million input tokens, outperforming GLM-5.1 in blind expert reviews. The model features 21 billion active …
Tencent released Hy3, a 295B Mixture-of-Experts model with Apache 2.0 weights, achieving a 78.0 SWE-bench Verified score and leading open-weight models on tool and search benchmarks. A free API on Ope…