# Hy3 Usage, Cost & Rank | OpenCode Data

> Hy3 ranked #30 by tokens across OpenCode last week, with 0.8% of tokens over the past two months. Hy3 costs $0.16 per 1M input tokens and $0.64 per 1M output tokens.

Tencent Hy reasoning model for coding, instruction following, and agent tasks

- Page: https://opencode.ai/data/tencent/hy3
- JSON: https://opencode.ai/data/tencent/hy3.json
- Model ID: tencent/hy3
- Lab: Tencent
- Updated: 2026-10-02T08:39:37.000Z

## Model facts

| Fact | Value |
| --- | --- |
| Context window | 256K |
| Max output | 128K |
| Knowledge cutoff | - |
| Release date | 2026-07-06 |
| Input modalities | text |
| Output modalities | text |
| Reasoning | Yes |
| Tool calling | Yes |
| Open weights | Yes |
| Weights | [Hugging Face](https://huggingface.co/tencent/Hy3) |

## Pricing: USD per 1M tokens

| Input | Output | Cached input | Cache write |
| --- | --- | --- | --- |
| $0.16 | $0.64 | $0.04 | $0.16 |

## OpenCode usage: past 2 months

| Metric | Value |
| --- | --- |
| Rank by tokens last week | #30 |
| Tokens | 6.9T |
| Share of all tokens | 0.8% |
| Change vs previous 2 months | +5269% |
| Unique users | 358K |
| Completed sessions | 6,097,909 |
| Average tokens per session | 1.1M |
| Average cost per session | $0.0129 |
| Total spend | $78,447 |
| Input tokens served from cache | 91.2% |
| Weekly retention | 68.1% |

## Daily usage: past 2 months

| Date | Tokens | Unique users | Sessions |
| --- | --- | --- | --- |
| 2026-08-08 | 4.8B | 1.4K | 1,729 |
| 2026-08-09 | 4.1B | 1.2K | 1,981 |
| 2026-08-10 | 5.6B | 1.6K | 2,740 |
| 2026-08-11 | 4.9B | 2.4K | 945 |
| 2026-08-12 | 125B | 7.7K | 146 |
| 2026-08-13 | 150B | 8.1K | 146 |
| 2026-08-14 | 148B | 8.1K | 58 |
| 2026-08-15 | 145B | 7.4K | 65 |
| 2026-08-16 | 138B | 7.1K | 78 |
| 2026-08-17 | 162B | 10K | 270 |
| 2026-08-18 | 167B | 9.8K | 53,712 |
| 2026-08-19 | 197B | 13K | 80,714 |
| 2026-08-20 | 276B | 18K | 123,177 |
| 2026-08-21 | 446B | 23K | 549,899 |
| 2026-08-22 | 364B | 17K | 487,323 |
| 2026-08-23 | 374B | 17K | 222,447 |
| 2026-08-24 | 473B | 22K | 351,588 |
| 2026-08-25 | 497B | 21K | 525,365 |
| 2026-08-26 | 579B | 25K | 824,432 |
| 2026-08-27 | 810B | 29K | 895,765 |
| 2026-08-28 | 752B | 25K | 776,711 |
| 2026-08-29 | 659B | 20K | 639,710 |
| 2026-08-30 | 126B | 8.8K | 138,429 |
| 2026-08-31 | 37B | 4.6K | 45,211 |
| 2026-09-01 | 27B | 3.8K | 33,272 |
| 2026-09-02 | 24B | 3.5K | 30,946 |
| 2026-09-03 | 19B | 3.2K | 17,207 |
| 2026-09-04 | 16B | 2.4K | 12,746 |
| 2026-09-05 | 11B | 1.8K | 10,139 |
| 2026-09-06 | 11B | 1.7K | 10,110 |
| 2026-09-07 | 13B | 2.3K | 20,696 |
| 2026-09-08 | 12B | 1.8K | 12,669 |
| 2026-09-09 | 11B | 1.6K | 10,845 |
| 2026-09-10 | 9.2B | 1.8K | 10,876 |
| 2026-09-11 | 6.3B | 1.5K | 7,761 |
| 2026-09-12 | 4.7B | 1.2K | 6,610 |
| 2026-09-13 | 4B | 1.1K | 7,024 |
| 2026-09-14 | 6.6B | 1.5K | 35,217 |
| 2026-09-15 | 6B | 1.6K | 18,225 |
| 2026-09-16 | 4.9B | 1.5K | 8,122 |
| 2026-09-17 | 5.4B | 1.6K | 9,851 |
| 2026-09-18 | 4.6B | 1.4K | 12,255 |
| 2026-09-19 | 3.5B | 1.1K | 8,754 |
| 2026-09-20 | 2.7B | 1.1K | 6,573 |
| 2026-09-21 | 4.2B | 1.3K | 5,179 |
| 2026-09-22 | 4.2B | 1.4K | 9,233 |
| 2026-09-23 | 3.8B | 1.2K | 6,067 |
| 2026-09-24 | 4B | 1.1K | 14,381 |
| 2026-09-25 | 3.3B | 920 | 13,962 |
| 2026-09-26 | 2.6B | 882 | 5,515 |
| 2026-09-27 | 1.9B | 906 | 4,060 |
| 2026-09-28 | 3.3B | 1.1K | 9,511 |
| 2026-09-29 | 3B | 1.1K | 6,063 |
| 2026-09-30 | 2.9B | 1K | 5,179 |
| 2026-10-01 | 2.7B | 922 | 5,290 |
| 2026-10-02 | 816M | 353 | 930 |

## Top countries: past 2 months

| Rank | Country | Tokens | Share |
| --- | --- | --- | --- |
| 1 | China | 1.4T | 20.9% |
| 2 | United States | 587B | 8.5% |
| 3 | India | 452B | 6.6% |
| 4 | Russia | 403B | 5.9% |
| 5 | Indonesia | 282B | 4.1% |
| 6 | Germany | 263B | 3.8% |
| 7 | Brazil | 249B | 3.6% |
| 8 | Singapore | 151B | 2.2% |
| 9 | Japan | 142B | 2.1% |
| 10 | France | 130B | 1.9% |
| 11 | Netherlands | 130B | 1.9% |
| 12 | Hong Kong | 118B | 1.7% |
| 13 | Vietnam | 115B | 1.7% |
| 14 | Spain | 107B | 1.6% |
| 15 | Egypt | 101B | 1.5% |

## Nearby models by tokens last week

| Rank | Model | Lab | Tokens |
| --- | --- | --- | --- |
| 26 | [kimi-k3](https://opencode.ai/data/moonshotai/kimi-k3.md) | Moonshot | 45B |
| 27 | [mimo-v2.5-pro](https://opencode.ai/data/xiaomi/mimo-v2-5-pro.md) | Xiaomi | 42B |
| 28 | [longcat-2.0](https://opencode.ai/data/meituan/longcat-2-0.md) | Meituan | 26B |
| 29 | [qwen3.8-max](https://opencode.ai/data/alibaba/qwen3-8-max.md) | Qwen | 26B |
| 30 | [hy3](https://opencode.ai/data/tencent/hy3.md) | Tencent | 17B |
| 31 | [minimax-m2.7](https://opencode.ai/data/minimax/minimax-m2-7.md) | MiniMax | 15B |
| 32 | [hy4-preview](https://opencode.ai/data/tencent/hy4-preview.md) | Tencent | 13B |
| 33 | [omen-alpha](https://opencode.ai/data/unknown/omen-alpha.md) | - | 12B |
| 34 | [grok-4.7](https://opencode.ai/data/xai/grok-4-7.md) | xAI | 9.6B |
| 35 | [jev-1.13](https://opencode.ai/data/unknown/jev-1-13.md) | - | 6B |

## Benchmarks

| Benchmark | Score | Metric | Source |
| --- | --- | --- | --- |
| SWE-Bench Verified | 78 | resolved | https://huggingface.co/tencent/Hy3 |
| SWE-Bench Multilingual | 75.8 | score | https://huggingface.co/tencent/Hy3 |
| SWE-Bench Pro | 57.9 | score | https://huggingface.co/tencent/Hy3 |
| Terminal-Bench | 71.7 | score | https://huggingface.co/tencent/Hy3 |
| NL2Repo | 45.6 | score | https://huggingface.co/tencent/Hy3 |
| DeepSWE | 28 | score | https://huggingface.co/tencent/Hy3 |
| BrowseComp | 84.2 | score | https://huggingface.co/tencent/Hy3 |
| WideSearch | 76.4 | score | https://huggingface.co/tencent/Hy3 |
| DeepSearchQA | 91 | score | https://huggingface.co/tencent/Hy3 |
| MCP Atlas | 79.1 | score | https://huggingface.co/tencent/Hy3 |
| Toolathlon | 48.5 | score | https://huggingface.co/tencent/Hy3 |
| APEX-Agents | 25.6 | pass@1 | https://huggingface.co/tencent/Hy3 |
| ClawEval | 68.5 | pass@3 | https://huggingface.co/tencent/Hy3 |
| WildClawBench | 53.6 | score | https://huggingface.co/tencent/Hy3 |
| SkillsBench | 55.3 | average over 3 runs | https://huggingface.co/tencent/Hy3 |
| Humanity's Last Exam | 53.2 | score | https://huggingface.co/tencent/Hy3 |
| Humanity's Last Exam | 37 | score | https://huggingface.co/tencent/Hy3 |
| GPQA Diamond | 90.4 | score | https://huggingface.co/tencent/Hy3 |
| FrontierScience | 21.3 | score | https://huggingface.co/tencent/Hy3 |
| FrontierScience | 74.8 | score | https://huggingface.co/tencent/Hy3 |
| USAMO | 72 | score | https://huggingface.co/tencent/Hy3 |
| MathArena Apex | 38.7 | score | https://huggingface.co/tencent/Hy3 |
| ArxivMath | 52.2 | score | https://huggingface.co/tencent/Hy3 |
| HorizonMath | 7.1 | pass@12 | https://huggingface.co/tencent/Hy3 |
| PHYBench | 77.4 | score | https://huggingface.co/tencent/Hy3 |
| CMT Benchmark | 37.8 | score | https://huggingface.co/tencent/Hy3 |
| IMOAnswerBench | 90 | score | https://huggingface.co/tencent/Hy3 |
| SuperChem | 54.9 | score | https://huggingface.co/tencent/Hy3 |
| CL-bench | 23.8 | score | https://huggingface.co/tencent/Hy3 |
| CL-bench-life | 17 | score | https://huggingface.co/tencent/Hy3 |
| AA-LCR | 73.4 | score | https://huggingface.co/tencent/Hy3 |

## Methodology

- Updates: Aggregated every hour. Days and weeks use UTC.
- Tokens: Input, output, reasoning, and cached tokens for each request.
- Users and sessions: Approximate counts of distinct users and OpenCode sessions.
- Cost: Session cost is the average cost per OpenCode session. Token prices are list prices from the OpenCode model catalog.
- Retention: The share of a model's users in one week who use it again the next week.
- Citation: Cite OpenCode Data (opencode.ai/data) with the update time shown at the top of the page.