impulse

Hy4 Preview

Open weightsPreview

Tencent: Hy4 preview is a mixture-of-experts model from Tencent, with 49B active parameters out of 770B total.

API id hy4-previewReleased 28 Aug 2026 (Aug 2026)
Input / 1M
$0.417
at Tencent TokenHub
Output / 1M
$1.25
at Tencent TokenHub
Context window
960K
tokens
Max output
64K
tokens per reply
Hosts
6
6 listings
Released
Aug 2026
28 Aug 2026

Pricing

List prices at Tencent TokenHub, in US dollars per million tokens.

TierInput / 1MOutput / 1M
Standard$0.417$1.25
Cached input
95% off input
$0.021

What it costs a month

At Tencent TokenHub list prices, 30 days.

  • Chat assistant5,000 requests a day, 2K in / 0.5K out, cached$183
  • RAG over documents2,000 requests a day, 8K in / 0.6K out, cached$150
  • Coding agent1,000 requests a day, 40K in / 2K out, cached$243
  • Overnight batch20,000 requests a day, 4K in / 0.4K out$1,301
Use your own numbers

Where to run it

6 hosts serve Hy4 Preview. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
hy4-preview
$0.417 / $1.25
hy4-preview
$0.834 / $2.50
+100% vs headline
DeepInfraInference
Hy4-preview
$0.834 / $2.50
+100% vs headline
GMI CloudInference
hy4-preview
$0.834 / $2.50
+100% vs headline
AIHubMixGateway
hy4-preview
$0.845 / $2.54
+103% vs headline
hy4-preview
$1.13 / $3.38
+170% vs headline

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
Text
Produces
Text
Features
Prompt caching
Open weights
License: Apache-2.0
Runs in about 1871.9 GB of GPU memory

Cheaper alternatives

Current models that cost less and keep most of what this one can do: tools, vision, reasoning, structured output, caching and enough context.

Same model, cheaper host

No host sells it for less than Tencent TokenHub.

Smaller Independent models

No cheaper Independent model keeps its key capabilities.

Other frontier labs

  • GPT-5.4 Nano
    OpenAI · on OpenAI
    $0.20 / $1.25
    Has every key capability
    26% less
  • Claude Haiku 5.5
    Anthropic · on Anthropic
    $0.10 / $0.50
    Has every key capability
    68% less
  • GPT-6 Luna
    OpenAI · on OpenAI
    $0.10 / $0.50
    Has every key capability
    68% less
  • GPT-4o Mini
    OpenAI · on OpenAI
    $0.15 / $0.60
    Lacks reasoning
    58% less

Open-weight picks

  • DeepSeek Flash
    DeepSeek · on DeepSeek
    $0.30 / $1.20
    Has every key capability
    16% less
  • Qwen3.7 Plus
    Alibaba Qwen · on Replicate
    $0.276 / $1.10
    Has every key capability
    23% less
  • Beam 1.0 501B A23B
    Independent · on W&B Inference
    $0.30 / $1.00
    Has every key capability
    24% less
  • MiMo V2.6 Pro
    Xiaomi · on GMI Cloud
    $0.348 / $0.696
    Has every key capability
    30% less

Model line

Newer and older versions, at each lab's own price.

  1. Aug 2026Hy4 PreviewLatest$0.417 / $1.25
  2. Apr 2026Hy3 Preview$0.18 / $0.60

Call it

Using Tencent TokenHub's API.

Python
from litellm import completion

response = completion(
    model="tencent/hy4-preview",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does Hy4 Preview cost?

On Tencent TokenHub, Hy4 Preview costs $0.417 per million input tokens and $1.25 per million output tokens. Cached input costs $0.021 per million tokens.

Which APIs offer Hy4 Preview?

Hy4 Preview runs on 6 APIs: Tencent TokenHub, OpenCode Go, DeepInfra, GMI Cloud, AIHubMix and Kyma API.

What is Hy4 Preview's context window?

960K tokens, with up to 64K tokens in one reply.

Is Hy4 Preview being retired?

No retirement date is scheduled for it on any API we track.

Is there a newer version of Hy4 Preview?

No. Hy4 Preview is the newest model in its line.

What are cheaper alternatives to Hy4 Preview?

Models that cost less and keep most of what it can do: GPT-5.4 Nano on OpenAI (26% less), Claude Haiku 5.5 on Anthropic (68% less) and DeepSeek Flash on DeepSeek (16% less). Savings are on a mix of 3 input tokens to 1 output token.

When was Hy4 Preview released?

Independent released it on 28 Aug 2026.

What is Hy4 Preview's API model id?

Use hy4-preview with Tencent TokenHub. Other hosts use their own ids: Hy4-preview on DeepInfra.

Sources

Raw catalog entry for hy4-preview on Tencent TokenHub
{
  "cache_read_input_token_cost": 2.1e-8,
  "comment": "These are the 50% promotional rates. Update on 2026-12-31 when the promotion ends at 23:59:59 Beijing time. List rates per 1M tokens are $0.834 input, $2.501 output and $0.042 cached input.",
  "input_cost_per_token": 4.17e-7,
  "input_cost_per_token_cache_hit": 2.1e-8,
  "litellm_provider": "tencent",
  "max_input_tokens": 960000,
  "max_output_tokens": 64000,
  "mode": "chat",
  "output_cost_per_token": 0.0000012505,
  "source": "https://www.tencentcloud.com/document/product/1300/78937",
  "supports_function_calling": true,
  "supports_prompt_caching": true,
  "supports_reasoning": true,
  "supports_response_schema": true
}