NOUMENOpen app
NOUMEN

tencent · Efficient

Tencent Hy3

Added Jul 6, 2026

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Hy3 is Tencent's 295B-parameter Mixture-of-Experts model with 21B active parameters, native 256K context, and configurable reasoning modes. It is built for coding, long-context comprehension, multi-turn dialogue, and agentic task execution with a focus on high-throughput production workloads.

  • Reasoning
  • Tools
  • Structured output
Input
Text
Output
Text
Reasoning effort
none · low · high
Discovery tier
Efficient

Pricing

USD per 1M tokens
Input
$0.066
Output
$0.26
Cache read
$0.029

Reference prices as published by the model provider. API example

Specifications

Context window
262.1K
Max output
128K
Average output (7 days)
1.4K tokens
Longer than 83% of 468 models

Side by side

Tencent Hy3 compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Tencent Hy3tencentThis modelGrok 4.20xAIRemove
ProvidertencentxAI
Discovery tierEfficientBalanced
AddedJul 6, 2026Mar 31, 2026
Limits
Context window262.14K2MBest
Max output128K131.07KBest
Average output1.4K tokens679 tokens
Pricing, USD per 1M tokens
Input$0.066Best$1.25
Output$0.26Best$2.50
Cache read$0.029Best$0.20
Performance
Throughput78.9 tok/s91.2 tok/sBest
Time to first token3.7 s2.9 sBest
Request success (24h)99.5%99.9%Best
Benchmark indexes
Intelligence Index25.325.7Best
Coding Index58.8—
Capabilities
ReasoningYesYes
ToolsYesYes
VisionNoYes
Structured outputYesYes
Reasoning effortnone, low, high—

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Tencent Hy3
  • Grok 4.20

Intelligence Index

  • Tencent Hy325.3
  • Grok 4.2025.7

Tencent Hy3: Better than 62% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Tencent Hy3: 58.8
    • Grok 4.20: —

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Tencent Hy3: 18.4%
    • Grok 4.20: —

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Tencent Hy3: 2.7%
    • Grok 4.20: —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Tencent Hy3: 947 Elo
    • Grok 4.20: —

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Tencent Hy3: 1,063 Elo
    • Grok 4.20: —

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Tencent Hy3: 11.2%
    • Grok 4.20: —

  • AA-LCRv1.1

    Reasoning across long documents

    • Tencent Hy3: 79.0%Top
    • Grok 4.20: 69.0%

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Tencent Hy3: 33.5%
    • Grok 4.20: 34.5%Top

  • GPQA Diamond

    Graduate-level science questions

    • Tencent Hy3: 89.7%
    • Grok 4.20: 91.1%Top

  • IFBench

    Following precise output instructions

    • Tencent Hy3: —
    • Grok 4.20: 81.2%

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Tencent Hy3: 0.5%
    • Grok 4.20: —

  • SciCode

    Scientific computing in Python

    • Tencent Hy3: 48.6%
    • Grok 4.20: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Tencent Hy3: —
    • Grok 4.20: 37.9%

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Tencent Hy3: —
    • Grok 4.20: 93.0%

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Tencent Hy3: 79.0%Top
    • Grok 4.20: 69.0%

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Throughput
78.9 tok/s
Faster than 52% of 254 models
Time to first token
3.7 s
Quicker than 48% of 254 models
Prompt cache hits
67.8%
About $0.0409 per 1M input tokens with caching
Request success (24h)
99.5%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 98.2%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tencent/hy3",
    "reasoning_effort": "none",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="tencent/hy3",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

tencent

Other tencent models, starting with the same family.