Hy4 Preview
The preview of Tencent's Hy4 generation, a 770B-parameter MoE (49B active) with open weights under Apache 2.0, aimed at coding, agents and office work.
- Params
- 770B, 49B active
- Context
- 1M tokens
- Max output
- 64K
- Released
- License
- Apache 2.0
Hosted API price
List price via Tencent · per 1M tokens
- Input
- $0.834 /1M
- Output
- $2.501 /1M
- Hosted by
-
2 providers
Hosted API pricing
How we price
Order
Providers are listed cheapest input rate first, then by output rate, a tie going to the creator's own listing. A provider that lists the model without a published rate sits at the end.
Prices
Base-tier, on-demand rates in USD per 1M tokens, converted at ECB reference rates where a provider publishes in another currency. Cached-input and batch rates show only where a provider publishes them. "Cheapest" marks the one provider with the lowest input and output pair; a tie wears no badge.
Cost
Input rate times the input tokens plus output rate times the output tokens, at the monthly volume set above the table. Cached-input, batch, long-context and reasoning-token billing are not modelled.
Transparency and funding
- Affiliates: Affiliate links are marked. We may earn a commission if you click them, but commissions never affect the order.
Every provider serving Hy4 Preview, cheapest input first. Set your monthly volume to see what each would bill.
| Provider | Input /1M | Output /1M | Cached input /1M | Cost at 10M in + 2M out | Link |
|---|---|---|---|---|---|
|
Tencent
Creator
|
$0.834 | $2.501 | $0.042 | $13.34 | Visit website |
|
Novita
|
$0.834 | $2.501 | $0.042 | $13.34 | Visit website |
Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.
Estimated cost to self-host
Hy4 Preview needs about 428 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× A100 at about $10,714 a month.
That costs the same as roughly 10B tokens a month on Tencent's API. Self-hosting is more expensive below that volume.
| Precision | Memory | Cheapest fit (32K context) | Cost /mo | Break-even vs API |
|---|---|---|---|---|
| 4-bit INT4 / FP4 | 428 GB |
8× A100
|
$10,714 | 10B tokens /mo |
| 8-bit FP8 / INT8 | 773 GB |
4× B300
|
$27,360 | 25B tokens /mo |
| 16-bit FP16 / BF16 | 1,545 GB |
8× B300
|
$54,720 | 49B tokens /mo |
Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. Break-even assumes a 5:1 input-to-output ratio. No guarantee of runtime support, usable performance, or that a matching quantized build exists. Pricing methodology.
Similarly priced models
The models nearest Hy4 Preview by blended rate, each at its own cheapest provider.
| Model | Blended /1M | Input /1M | Output /1M | Context | Cutoff | vs Hy4 Preview |
|---|---|---|---|---|---|---|
|
Nemotron 3 Ultra 550B A55B
Nvidia
|
$0.99 | $0.66 | $2.64 | 1M | Sep 2025 | −11% |
DeepSeek R1
DeepSeek
|
$1.00 | $0.70 | $2.50 | 128K | −10% | |
|
DeepSeek R1 Turbo
DeepSeek
|
$1.00 | $0.70 | $2.50 | 64K | −10% | |
Llama 3 70B
Meta
|
$1.00 | $0.65 | $2.75 | 8K | Dec 2023 | −10% |
Kimi K2.6
Moonshot AI
|
$1.0333 | $0.60 | $3.20 | 262K | −7% | |
|
Hy4 Preview
This model
Tencent
|
$1.1118 | $0.834 | $2.501 | 1M | ||
Grok Build 0.1
xAI
|
$1.1667 | $1.00 | $2.00 | 256K | +5% | |
|
DeepSeek V4 Pro
DeepSeek
|
$1.2117 | $0.91 | $2.72 | 1M | +9% | |
GLM-5.2
Z.AI
|
$1.2167 | $0.90 | $2.80 | 1M | +9% | |
|
GLM-5.1
Z.AI
|
$1.2377 | $0.825 | $3.301 | 205K | +11% | |
Gemini 3.6 Flash
Google Cloud
|
$1.25 | $0.75 | $3.75 | 1M | Mar 2026 | +12% |
USD per 1M tokens at each model's cheapest listed provider. Blended is the cost of 10M input plus 2M output tokens, spread over the 12M.
Capabilities
What Hy4 Preview accepts and can do, as published by Tencent.
- Text Accepts and generates natural-language text.
- Reasoning Thinks before it answers, always on or as a switchable mode.
- Function calling Connect to external tools, APIs, and systems.
- Structured output Return responses in structured formats like JSON.
Common questions
What is Hy4 Preview good for?
Coding, agent tasks and office work such as turning scattered files into documents, spreadsheets and decks, with reasoning modes from no-think up to high. Apache 2.0 weights allow commercial self-hosting, on a multi-GPU node.
When is Hy4 Preview not a good fit?
Image or video input: it is text only. Tencent calls it an early version that can reason longer than needed and over-check its own work, and self-hosting takes a large multi-GPU node.
What is the cheapest way to run Hy4 Preview?
Hosted, unless you push serious volume. Tencent charges $0.834 in / $2.501 out per 1M tokens. The cheapest rental that fits is 8x A100 at $10,714 a month, which costs the same as about 10B tokens a month on that API.
How does Hy4 Preview compare with Hy4?
Tencent says the full Hy4 improves on Hy4 Preview in instruction following, long-context comprehension and long-horizon agent tasks, hallucinates less in its internal tests, and adds visual understanding for the first time. Hy4 Preview lists at 100% more per 1M input tokens and 100% more per 1M output tokens than Hy4. It came out 3 months earlier, has a larger context (1M against 1M tokens), takes no image input, which Hy4 does and publishes its weights under Apache 2.0, which Hy4 does not.
Can I self-host Hy4 Preview?
Yes. The weights are Apache 2.0 licensed. At 4-bit it needs about 428 GB of GPU memory, which starts at roughly $10,714 a month on the cheapest rental that fits.