Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort:...
Tested on 9 benchmarks. Top scores: Chatbot Arena Elo — Coding (1507.5%), Chatbot Arena Elo — Overall (1456.0%), Artificial Analysis · GPQA Diamond (89.7%).
Chatbot Arena coding Elo. Human preference ranking specifically for coding tasks and technical questions.
Chatbot Arena overall Elo rating. Crowdsourced human preference ranking from blind head-to-head comparisons across all topics.
- Typetext
- Context262K tokens (~131 books)
- ReleasedJul 2026
- LicenseProprietary
- StatusActive
- Cost / Message~$0.001
Frequently Asked Questions
Related Models
Qwen3 Coder Next FP8 · AlibabaQwen2.5 Coder 32B Instruct AWQ · AlibabaSpeaker Diarization Community 1 · PyannoteGemma 3 12B (free) · Google DeepMindQwen3 Next 80B A3B Instruct (free) · Alibaba QwenKey facts · as of 2026-10-05
- Hy3 by tencent. Not enough public benchmark scores to rank yet.
- List price $0.13 input · $0.53 output per 1M tokens (as of 2026-10-05).
- Sold by 6 providers (as of 2026-10-05): DeepInfra (fp4) $0.13 in / $0.53 out · Tencent (fp8) $0.13 in / $0.53 out · GMICloud (bf16) $0.14 in / $0.58 out · Novita $0.14 in / $0.58 out · Phala $0.15 in / $0.64 out · and 1 more. Every provider
How to cite · data as of 2026-10-05
Hy3 · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/hy3
Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP