Tencent Hunyuan Hy3 Launched: 295B MoE Model for Coding and Agents
AGI Hunt · wechat · 2026-07-06
Tencent's Hunyuan Hy3 model is officially launched. It is a MoE model combining fast and slow thinking, featuring a total of 295B parameters, 21B activated parameters, and a 256K context length, primarily targeting Coding and Agent scenarios. Official benchmarks indicate its Coding capabilities are on par with DeepSeek V4 Pro and Seed2.1 Pro. It leads domestic models in Agent tasks (browsing, MCP, Skills), trailing only slightly behind Claude Opus 4.8 on certain metrics. Pricing is set at 1 yuan per million input tokens, 4 yuan per million output tokens, and 0.25 yuan for cached input hits.
In a hands-on test using Tencent Desktop Agent WorkBuddy (the first to integrate Hy3 with a two-week limited free trial), the author had it generate a 23-page PPT on 'History of Western Art' and a native macOS guitar tuner app. The author praised the stunning results and fast speed, recommending that non-programmers prioritize the out-of-the-box GUI format WorkBuddy.
Related event: Tencent Open-Sources Hunyuan Hy3 for Agentic Workloads(26 posts)→
More from coding & agent
- agents-best-practices: a provider-neutral Agent Skill for designing and auditing agentic harnesses — tom_doerr · 2026-09-11
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11
- First-ever Three.js Conference lands in Paris, with a panel on AI-shortened design workflows — OdinLovis · 2026-09-11
- Data engineering, not agent frameworks, is the real bottleneck for enterprise AI agents — dhruv2038 · 2026-09-11
- RTK Terminal Compression Cuts Tokens but Leaves Your AI Coding Bill Unchanged — Bartaseth · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11