Hy3
- In Input modalities: Text Out Output modalities: Text
- Reasoning· low, high, none
- Reasoning can be off
- Tool use
- Structured output
- Open weights
Prices
1 offering · prices as listed| Conditions/ 1M tokens | Input | Cache read | Output |
|---|---|---|---|
| Standard | ¥1.00 | ¥0.25 | ¥4.00 |
| Batch | ¥0.50 | ¥0.125 | ¥2.00 |
Prices from 腾讯云 TokenHub：模型价格 (opens in a new tab) (checked Oct 30, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Max input
- 192K
- Licence
- Apache 2.0
- Family
- hy
- Aliases
- hy3-preview, hy3-202608
- Model ID
Not stated on Tencent’s pages: max thinking budget, built-in tools, knowledge cutoff.
Run locally
Weights: tencent/Hy3- Parameters 298.8B
- Active 21B · 192 experts, 8 per token
- Native precision BF16
- Layers 80
- KV cache per token 320 KB
- Config positions 256K (config.json max_position_embeddings)
- API context 250K (maker’s API)
- Needs
- ≈ 157.7 GB (weights 147.8 + KV cache 2.5 + overhead 7.4)
- Usable
- 21.6 GB (90% of its memory)
- Try
- IQ2_XXS runs (needs system memory), but below 4 bits quality drops noticeably
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 556.5 GB | ≈ 586.9 | ✕Doesn’t fit | — |
| Q8_0 | 295.7 GB | ≈ 312.9 | ✕Doesn’t fit | — |
| FP8maker’s file | 279.3 GB | ≈ 295.8 | ✕Doesn’t fit | — |
| Q6_K | 228.5 GB | ≈ 242.5 | ✕Doesn’t fit | — |
| Q5_K_M | 197.9 GB | ≈ 210.3 | ✕Doesn’t fit | — |
| Q5_0 | 191.3 GB | ≈ 203.4 | ✕Doesn’t fit | — |
| Q4_K_M | 169.0 GB | ≈ 180.0 | ✕Doesn’t fit | — |
| Q4_0 | 156.5 GB | ≈ 166.9 | ✕Doesn’t fit | — |
| IQ4_XS | 147.8 GB | ≈ 157.7 | ✕Doesn’t fit | — |
| Q3_K_Mquality drops | 136.0 GB | ≈ 145.3 | ✕Doesn’t fit | — |
| IQ3_XXSquality drops | 106.4 GB | ≈ 114.3 | ✕Doesn’t fit | — |
| Q2_Kquality drops | 104.3 GB | ≈ 112.1 | ✕Doesn’t fit | — |
| IQ2_XXSquality drops | 71.7 GB | ≈ 77.7 | ○Needs system memory 10–15 tok/s | — |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): no single listed GPU; unified memory Apple M5 Ultra (64-core GPU) · 256 GB.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by us- LMArena Text 1,455
#50 of 163 listed models · CI 1,449–1,460 · 13,840 votes
- LMArena Text · Chinese 1,497
#46 of 161 listed models · CI 1,479–1,516 · 1,065 votes
- LMArena Text · Coding 1,503
#49 of 163 listed models · CI 1,493–1,513 · 3,772 votes
- LMArena Text · Creative writing 1,426
#52 of 163 listed models · CI 1,414–1,438 · 2,750 votes
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Oct 28, 2026
- changes we made
Changes
All changes →- 3 new prices on Tencent Cloud TokenHub, China Input (Batch) ¥0.50 · Output (Batch) ¥2.00 · Cache read (Batch) ¥0.125
Sources
- 腾讯云 TokenHub：模型价格in Chinese (opens in a new tab) cloud.tencent.com · checked Oct 30, 2026
- 腾讯云 TokenHub：混元调用指南in Chinese (opens in a new tab) cloud.tencent.com · checked Nov 2, 2026
- tencent/Hy3 on Hugging Face (Tencent) (opens in a new tab) huggingface.co · checked Nov 1, 2026
- Tencent Hy3 Now Available Globally, Extending Practical AI Across Products, Workflows and Cloud Services - Tencent (opens in a new tab) tencent.com · checked Nov 1, 2026
- 模型列表in Chinese (opens in a new tab) intl.cloud.tencent.com · checked Oct 31, 2026
- 大模型服务平台 TokenHub 模型列表_腾讯云in Chinese (opens in a new tab) cloud.tencent.com · checked Nov 1, 2026
- 大模型服务平台 TokenHub 深度思考_腾讯云in Chinese (opens in a new tab) cloud.tencent.com · checked Nov 1, 2026
- 大模型服务平台 TokenHub Token Plan 个人版积in Chinese (opens in a new tab) cloud.tencent.com · checked Nov 1, 2026
- Tencent Hunyuan Officially Releases Hy3, Advancing Agent Capabilities and Deeper Product Integration - Tencent (opens in a new tab) tencent.com · checked Nov 5, 2026
- tencent/Hy3 model card (opens in a new tab) huggingface.co · checked Nov 1, 2026
- tencent/Hy3 model card (opens in a new tab) huggingface.co · checked Oct 30, 2026
- tencent/Hy3 config.json (opens in a new tab) huggingface.co · checked Nov 2, 2026
- tencent/Hy3 files (opens in a new tab) huggingface.co · checked Nov 1, 2026
- tencent/Hy3-FP8 files (opens in a new tab) huggingface.co · checked Oct 30, 2026
- tencent/Hy3 README.md (opens in a new tab) huggingface.co · checked Oct 31, 2026
Tencent · China · tencent.com
Related models
Snapshots, the same family and other models from this provider