Private AI

Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Nano GPT logo
NanoGPT
Search
 
 
 
 
All conversations

Create

Chat
Studio

Explore

Models
Pricing
AI Apps
Blog

Library

Gallery
Workspace
Usage
API
Suggestions and bugs
Use cases

|

Terms

|

Privacy

|

Refunds
Balance
Subscription
Teams
Referrals
Presets
Settings
FAQPress & mediaBrand assetsBenchmarksEarn crypto
  1. Models
  2. Text
  3. Tencent

Tencent Hy3

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
  • Tool Calling
  • Structured Output

Added Jul 6, 2026

Model weights

Pricing

Auto routing · per 1M tokens
Input
$0.066
Output
$0.26
Cache read
$0.029
Compare provider prices

Specifications

Context window
262.1K
Max output
262.1K
Parameters
295B / 21B
Total / active
Avg output (7d)
1.5K tokens
Longer than 86% of models

Benchmarks

Sourced from Artificial Analysis.

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

18.4%

Better than 36% of models compared

AutomationBench-AA Tasks Completed

Fully completed workflows without guardrail violations

2.7%

Better than 36% of models compared

AA-Briefcase v1.1

Agentic knowledge work (Elo)

947 Elo

Better than 41% of models compared

GDPval-AA v2.1

Economically valuable tasks (Elo)

1071 Elo

Better than 47% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

11.2%

Better than 43% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

79.0%

Better than 80% of models compared

Reasoning

HLE

Humanity's Last Exam

33.5%

Better than 79% of models compared

IFBench

Instruction-following benchmark

63.1%

Better than 74% of models compared

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

0.5%

Better than 32% of models compared

SciCode

Python programming for scientific computing

48.6%

Better than 46% of models compared

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

89.7%

Better than 88% of models compared

Terminal-Bench Hard (legacy)

Agentic coding and terminal use

34.1%

Better than 79% of models compared

T²-Bench Telecom (legacy)

Conversational AI agents in dual-control scenarios

92.7%

Better than 88% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

79.0%

Better than 80% of models compared

Last updated Nov 4, 2026

Artificial Analysis

Providers

Provider information for this model’s automatic routing. These routes cannot be selected individually.

Loading provider options…

Related text models

Compare Tencent Hy3 with similar models from the same provider or model family.

Tencent Hy4 Preview

tencent/hy4-preview

Hy4 Preview is Tencent's 770B-parameter mixture-of-experts model with 49B active parameters. It is designed for coding agents, complex tool-use workflows, and productivity tasks, with a 1M-token context window and configurable reasoning effort.

Assistant Pepe 32B

sicariussicariistuff/assistant-pepe-32b

Assistant Pepe 32B is SicariusSicariiStuff's Qwen3-32B finetune for creative writing, brainstorming, roleplay and irreverent conversation. Trained on UBW Tapestries, it has a built-in conversational persona and replies without a separate thinking mode. It accepts and returns text, with a 40,960-token context window and up to 8,192 output tokens.

OrcaCyber MiniMax M3.1 Flash Uncensored

orca/orcacyber-minimax-m3.1-flash

An abliterated / uncensored variant of MiniMax M3.1 Flash for authorized security research, coding, and tool use. Reasoning is always enabled, with selectable effort. Supports a 262K-token context window, up to 131K output tokens, tool calling, and structured output. Accepts text input. Offensive work is permitted only within an authorized security engagement.

OrcaCyber Qwen3.8 Flash Next Uncensored

orca/orcacyber-qwen3.8-flash-next

An abliterated / uncensored variant of Qwen3.8 Flash Next for authorized security research, coding, and tool use. Accepts text and images, with reasoning enabled by default and an option to turn it off. Supports a 262K-token context window, up to 131K output tokens, tool calling, and structured output. Offensive work is permitted only within an authorized security engagement.

GLM 5.3 FlashX Legal

z-ai/glm-5.3-flashx-legal

GLM 5.3 FlashX Legal is Z.AI's model trained for legal document analysis, contract review, legal reasoning and initial drafts. It accepts text and images and returns text, with always-on reasoning, tool calling, JSON output, a 1M-token context window and up to 131,072 output tokens. External legal databases and document-editing tools require your own integrations.

DeepSeek V4.1 Flash Heretic

pqhaz/DeepSeek-V4.1-Flash-1017-heretic

pqhaz's Heretic checkpoint of DeepSeek V4.1 Flash 1017, served by Sable at original mixed precision (FP8 dense/Engram and MXFP4 experts). Supports text and image input, tools, structured output, and optional reasoning. Separate from the DeAlign uncensored checkpoint.