What hardware do you need to run Tencent Hy3?
Tencent Hy3 at Q4_K_M is 182 GB of weights, with a context ceiling of 256k tokens. Tencent's Apache-licensed flagship. Needs an Ultra-class machine.
How good is Tencent Hy3, really?
On the Artificial Analysis Intelligence Index v4.3 it scores 26, which puts it in the Haiku-class band. In the same band as Anthropic's cheap, fast tier. Every current OpenAI model scores above this band. See the whole table.
Head to head with its neighbours on the leaderboard: vs Inkling Small above it, vs Ling 3.0 flash below. One more model needs much the same memory at 32k of context: vs GLM-5.3-Flash.
What Tencent Hy3 costs either way
Renting the same model costs $0.0825 per million input tokens and $0.33 per million output (OpenRouter, cheapest active endpoint, checked 2026-10-29). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill. What a million tokens costs each way puts the two prices side by side.
Tencent Hy3 fits at 16k of context
Every figure above is taken at 32k, where Tencent Hy3 needs 193 GB. The weights are 182 GB of that and they do not move; the rest is the key-value cache, which grows with every token you keep. Ask for 16k instead and the cache falls from 11 GB to 5.4 GB, so the model needs 188 GB, which the Mac Studio M5 Ultra, 256GB holds in its 192 GB. At $10,799 it runs at 20 tok/s and pays back in 1,031 years at 500k tokens a day.
Run Tencent Hy3 at 16k on the Mac Studio M5 Ultra, 256GB
The machines that miss Tencent Hy3, and what they run
Of the machines here with a published price that are still sold, every one but the Mac Studio M5 Ultra, 256GB, which holds it at 16k, misses Tencent Hy3, and none of them misses by a window: its 182 GB of weights are larger than the usable memory in every one of them, before a single token of key-value cache. The DGX Spark GB10 Grace Blackwell, 128GB comes nearest, 64 GB short at 4k, the shortest window on the list. Qwen3.8 27B is the strongest model seven of the eight hold, and it scores 34 on the intelligence index against this model's 26.
Two machines here do hold Tencent Hy3 at 32k, and neither of them is an answer to what to buy, which is the question the rest of this page asks: the Mac Studio M3 Ultra, 512GB, no longer sold, at the $9,499 it launched at, and the Mac Studio M5 Ultra, 512GB, on sale with no published price. They are sold with 512 GB of memory and hand a model 384 GB of it, against the 193 GB Tencent Hy3 needs at 32k.
| Machine | Price | Usable memory | Short by | Strongest model it holds |
|---|---|---|---|---|
| DGX Spark GB10 Grace Blackwell, 128GB | $4,699 | 119.5 GB | 64 GB | Qwen3.8 27B |
| Strix Halo Framework Desktop, 128GB | $3,449 | 96 GB | 87.5 GB | Qwen3.8 27B |
| Mac Studio M5 Max, 128GB | $5,099 | 96 GB | 87.5 GB | Qwen3.8 27B |
| MacBook Pro M5 Max (16-inch), 128GB | $6,999 | 96 GB | 87.5 GB | Qwen3.8 27B |
| NVIDIA RTX PRO 6000 Blackwell, 96GB | $18,000card only | 95 GB | 88.5 GB | Qwen3.8 27B |
| Mac mini M5 Pro, 64GB | $2,699 | 48 GB | 135.5 GB | Qwen3.8 27B |
| AMD Radeon AI PRO R9700, 32GB | $1,299card only | 31 GB | 152.5 GB | Qwen3.8 27B |
| MacBook Air M5 (13-inch), 16GB | $1,299 | 10.5 GB | 173 GB | Gemma 4 12B |
One machine per family, the largest configuration in each that still misses, nearest first. Short by is the weights plus the key-value cache at 4k measured against the memory that machine's GPU can address, which that machine's own page gives. The last column is the highest-scoring current model the machine holds at 32k.
Price the DGX Spark GB10 Grace Blackwell, 128GB on Qwen3.8 27B instead
The specifics
- Parameters
- 295B, of which 21B are active per token
- Quantisation
- Q4_K_M
- Weights on disk
- 182 GB
- KV cache
- 11 GB at 32k context — Plain grouped-query attention on all 80 layers. How weights and cache add up.
- Maximum context
- 256k tokens
- Licence
- Apache 2.0
- Sources
- tencent/Hy3