Hy3 preview
2976°
Run this model
Offers / Ads
Tencent | 262,144 token context | $0.18/M input, $0.60/M output. Hy3 preview is a high-efficiency Mixture-of-Experts model from Tencent designed for agentic workflows and production use. It supports configurable reasoning levels across disabled, low, and high modes, allowing it to balance speed and depth depending on the task, while delivering strong code generation and reliable performance across multi-step, real-world workflows.
Guides for this model
Best local LLM by use case (October 2026): the right open model for every job
How much VRAM do I actually need for local AI?
Open weights vs open source vs API-only: what you are actually allowed to do
What is a GGUF file? Quantized models explained (Q4, Q8, fp8, NF4)
Captain's notes
Notes from 2026-10-13
Hy3 preview is a 295B-parameter Mixture-of-Experts model from Tencent, activating only 21B parameters per forward pass. It ships with configurable reasoning depth—off, low, or high—so you can trade cost for thoroughness on a per-task basis. The weights are open, the price is low, and it is built explicitly for agentic and multi-step workflows rather than casual chat.
Known for
- Configurable reasoning modes let you dial speed versus depth per request.
- Strong coding and agent benchmark results, including SWE-bench Verified.
- 256K token context window handles long documents and multi-turn agent loops.
- MoE architecture keeps active parameters at 21B despite 295B total scale.
- Open weights available on Hugging Face; deployable via vLLM or SGLang.
Watch out for
- 295B total parameters means self-hosting demands serious hardware despite the lean active count.
- Still a preview release; Tencent has not published a final licence or stability guarantees.
- Reasoning mode adds output tokens, which raises cost when set to high.
Prompts to try
Multi-step refactor task
You have a Python service that reads a CSV, filters rows where the 'status' column equals 'active', computes the mean of the 'score' column per 'region', and writes the result to a new CSV. The current code uses a nested loop and is slow on files over 100 MB. Rewrite it using pandas, add error handling for missing columns, and explain each change.
Agent planning under constraints
You are an agent with three tools: web_search, read_file, and write_file. A user wants a competitive analysis of the top three open-weight LLMs released in 2025, saved as a markdown table comparing context length, licence, and best use case. Write out the exact sequence of tool calls you would make, with arguments, before executing anything.
Hard reasoning puzzle
Five engineers each own exactly one of five different laptops and prefer exactly one programming language. The Rust developer sits next to the person with the ThinkPad. The Python developer owns the MacBook. The person with the Dell is not at either end. The Go developer is immediately to the left of the C++ developer. The TypeScript developer owns the XPS. Who owns the Surface, and what language do they use? Show your reasoning step by step.
Compare with
- Hy4 previewOpen-weights model1,048,576 token context · $0.83/M input, $2.50/M output
- Hy3Open-weights model262,144 token context · $0.13/M input, $0.53/M output
- Hunyuan A13B InstructOpen-weights model131,072 token context · $0.14/M input, $0.57/M output
- GPT-6 LunaAPI-only model1,050,000 token context · $0.10/M input, $0.50/M output
More text models from tencent
Questions people ask
- Are the Hy3 preview weights open and can I run it locally?
- Yes, the weights are publicly available on Hugging Face under tencent/Hy3-preview, and Tencent documents deployment via vLLM and SGLang. Bear in mind the model is 295B parameters total, so local inference requires substantial hardware.
- What does Hy3 preview cost and what is its context length?
- Via API it is priced at $0.18 per million input tokens and $0.60 per million output tokens, with a 262,144-token context window.
- How does Hy3 preview differ from DeepSeek-V3 or Kimi-K2?
- All three are large MoE models, but Hy3 preview activates only 21B parameters versus 37B for DeepSeek-V3 and 32B for Kimi-K2, and it adds selectable reasoning depth modes that neither of those models expose as a first-class control.
Captain's verdictA genuinely lean MoE with open weights, low API prices, and reasoning you can actually tune—worth a serious look for agent and coding work.
Model card on Hugging Face
Captain writes these notes from the model's own description and model card. Prices and context sizes are the provider's at the time of writing and can change.
Ranking badge
Updated Oct 22, 2026
Made this model? Show this badge on your page.
Paste it into your Hugging Face card or Civitai description.
How this score adds up
This model has no popularity data from any source yet, so it carries a stand-in number below 7,000° that keeps the lists complete. It never ranks above a model with data. The breakdown appears here as soon as a source reports it.
Glossary: the terms used here
- Heat score
- One number for how much a model is used and wanted right now, from 0° to 9,999°. Every list on this site that says hottest is sorted by it and nothing else.
- Floor
- The 7,000° every model with real data starts from. The scale runs from 0° to 9,999°; the part below the floor is reserved for stand-ins (models without data), so that a model with data always ranks above one without. The floor is the same for everyone and decides nothing; the degrees above it are the measurement.
- Download-equivalents
- Everything the score treats as a download, added up: downloads themselves, each torrent seeder as 1,000, each AI Horde image and each run on a generation site as one.
- Log scale
- Counting by tens, not by ones: going from 1,000 to 10,000 downloads adds the same degrees as going from 10,000 to 100,000. It keeps giants from crowding out everything else.
- Likes
- Thumbs up and likes from every source, worth half a download each on the log scale.
- Adoption rate
- How many downloads a day a model gets right now, from the difference between two readings of a source. A quarter of the weight of the download term.
- Measured and estimated
- Measured means two readings of a source exist and the difference is real. Estimated means only today's total is known and the past is spread over the model's life; the site says which one it is showing.
- Seeder
- Someone who keeps a copy of the weights online for others to download over a torrent, or a hosted mirror doing the same. Each one counts as 1,000 downloads.
- AI Horde
- A volunteer network that generates images for free on donated computers. The number of images it has made with a model is public, and each counts as one download.
- Generation sites
- Sites that host their own copies of a model and let people generate with it, such as PixAI, SeaArt and Yodayo. Their downloads count like Civitai's and each generation run as one download.
- OpenRouter tokens
- A text model served through an API, such as Claude or GPT, has no weights to download, so its usage is measured in tokens: the pieces of text it read and wrote through OpenRouter, a gateway many apps use, over the last 30 days. Every 10,000,000 tokens count as one download. Source: OpenRouter (openrouter.ai/rankings), CC BY 4.0.
- Full model (checkpoint)
- The weights themselves, the file that generates on its own. It gets a bonus because add-ons depend on it.
- LoRA and embedding
- Small add-on files that steer someone else's weights toward a style, a character or a concept. They cannot run alone.
- Family and weights
- The base a model is built on, such as SD 1.5, SDXL, Flux or Wan. Every model inherits the release date of its family's weights.
- Age
- Degrees taken off for how long ago the family's weights came out: 150° a year, up to 750°. It keeps old downloads that had years to pile up from freezing the top of the list.
- Family boost
- A small, bounded nudge for LoRAs of a family this site is pushing forward right now, always listed in the breakdown.
- Stand-in score
- A stable number below the floor given to a model no source reports on yet, so lists stay complete. It never ranks above a model with data.
- New! (grace period)
- A model less than 7 days old shows New! instead of a score and holds no place in the rankings. Its score is computed from day one, but a model one day old grows faster than anything, so without the wait the week's #1 would always be whatever came out yesterday rather than what held up for a week.
- Trend
- How a model's growth over a period compares with the typical model's growth over the same period. Faster than typical reads as a rise, slower as a fall.
- Typical model
- The middle of the catalog: half the models with a figure grew faster over the period, half slower.
Hy3 preview is a text model for the Text / LLM family listed on ModelTrends.ai, a read-only catalog of open source AI image, video and text models. Compare it with other Text / LLM models and check its heat score to see how it is trending.
Browse by Model Family
Text / LLM Models
434
Anima Models
528
Chroma Models
19
Fibo Models
2
Flux Models
1,496
Flux 2 / Klein Models
1,239
Flux 3 Models
3
Moss Models
3
Iris Models
1
MiniMax H3 Models
364
Hunyuan Models
204
Ideogram Models
20
Krea2 Models
717
Qwen Models
20
Qwen2 Models
47
Ltx2 Models
153