Hy3
Hy
Tencent Hy reasoning model for coding, instruction following, and agent tasks
ReasoningTool CallingOpen Weights
Context Window
256K
Max Output
128K
Temperature
Yes
Open Weights
Yes
Knowledge Cutoff
N/A
Released
2026-07-06
Last Updated
2026-07-06
Modalities
Input:Text
→Output:Text
Benchmarks
| Benchmark | Score | Metric | Harness |
|---|---|---|---|
| SWE-Bench Verified | 78 | resolved | - |
| SWE-Bench Multilingual | 75.8 | score | SWE-agent |
| SWE-Bench Pro | 57.9 | score | SWE-agent |
| Terminal-Bench | 71.7 | score | Terminus 2 |
| NL2Repo | 45.6 | score | Claude Code |
| DeepSWE | 28 | score | mini-swe-agent |
| BrowseComp | 84.2 | score | Tencent internal search harness |
| WideSearch | 76.4 | score | Tencent internal search harness |
| DeepSearchQA | 91 | score | Tencent internal search harness |
| MCP Atlas | 79.1 | score | Scale AI |
| Toolathlon | 48.5 | score | - |
| APEX-Agents | 25.6 | pass@1 | - |
| ClawEval | 68.5 | pass@3 | Tencent internal harness |
| WildClawBench | 53.6 | score | OpenClaw |
| SkillsBench | 55.3 | average over 3 runs | Claude Code |
| Humanity's Last Exam | 53.2 | score | - |
| Humanity's Last Exam | 37 | score | - |
| GPQA Diamond | 90.4 | score | - |
| FrontierScience | 21.3 | score | - |
| FrontierScience | 74.8 | score | - |
| USAMO | 72 | score | - |
| MathArena Apex | 38.7 | score | - |
| ArxivMath | 52.2 | score | - |
| HorizonMath | 7.1 | pass@12 | - |
| PHYBench | 77.4 | score | - |
| CMT Benchmark | 37.8 | score | - |
| IMOAnswerBench | 90 | score | - |
| SuperChem | 54.9 | score | - |
| CL-bench | 23.8 | score | - |
| CL-bench-life | 17 | score | - |
| AA-LCR | 73.4 | score | - |
Available Providers (9)
| Provider | Input /1M | Output /1M | Cache Read /1M | Cache Write /1M | Reasoning | Status |
|---|---|---|---|---|---|---|
 NanoGPT | Rp 1.181 | Rp 4.649 | Rp 519 | — | effort | — |
 LLM Gateway | Rp 2.361 | Rp 9.441 | Rp 591 | — | effort | — |
 Kilo Gateway | Rp 2.361 | Rp 9.441 | Rp 591 | — | effort | — |
 Novita AI | Rp 2.504 | Rp 10.371 | Rp 626 | — | effort | — |
 OpenRouter | Rp 2.361 | Rp 9.441 | Rp 591 | — | effort | — |
 Vercel AI Gateway | Rp 2.504 | Rp 10.371 | Rp 626 | — | effort | — |
 OrcaRouter | Rp 3.219 | Rp 10.550 | Rp 1.055 | — | effort | — |
 Eden AI | Rp 2.361 | Rp 9.441 | Rp 591 | — | effort | — |
 CrossModel | Rp 2.861 | Rp 11.444 | Rp 716 | Rp 2.861 | effort | — |