GLM (Z.ai)
Open-weight frontier LLM family from Z.ai (Zhipu AI), tuned for coding and agents.

Tencent's open and hosted AI models for text, image, and video
Hunyuan is one of the more developer-friendly Chinese model families thanks to its aggressive open-sourcing and class-leading open video generation, plus very low API pricing. The trade-offs are a sprawling, sometimes confusing model lineup, vendor-sourced benchmarks, a China-centric consumer app, and the data-residency and regulatory concerns that come with any Chinese cloud AI for Western enterprises.
Tencent Hunyuan is a broad AI model family covering text, image, 3D, and video, delivered through the Yuanbao consumer app, the Tencent Cloud API, and open-weight releases. It is best known for open-sourcing high-quality video generation (HunyuanVideo) and for very low per-token pricing, making it a strong pick for cost-sensitive developers and creators, with the usual caveats about vendor benchmarks and China-cloud data residency.
Tencent Hunyuan is Tencent's in-house foundation-model suite, covering large language models plus image, 3D, and video generation. It stands out for how much it open-sources: several models, including the HunyuanVideo generator and the HY3 text model, have been released publicly, some under the permissive Apache 2.0 license, alongside very low per-token pricing for hosted API access. On the text side, the HY3 model (reported in 2026) is a Mixture-of-Experts design with a 256K-token context window and selectable inference modes, positioned as a cost-efficient option for agentic and coding tasks. On the generative-media side, HunyuanVideo and its image-to-video and higher-resolution successors have drawn attention for cinematic quality and bilingual (Chinese and English) prompting. As with other Chinese labs, benchmark and parameter claims come largely from Tencent and third-party aggregators and should be read as claims. Hunyuan reaches users through three channels: the consumer Yuanbao assistant app, the Tencent Cloud API for developers and enterprises, and open-weight releases on Hugging Face and GitHub for self-hosting. This makes it attractive to developers and creators who want free, high-quality open video and image generation, though the consumer experience is China-centric and Western enterprises face the usual data-residency and regulatory considerations.
Tencent Hunyuan is Tencent's suite of AI models for text, image, 3D, and video, delivered via the Yuanbao app, Tencent Cloud API, and open-weight releases. It is best known for open-sourcing strong video generation (HunyuanVideo) and for very low per-token API pricing. The lineup is broad and fast-moving, and benchmark claims are vendor-sourced. It suits cost-conscious developers and creators, with China-cloud caveats for Western enterprises.
Hunyuan is developed by Tencent Holdings Ltd. (HKEX: 0700), one of the world's largest technology companies, with businesses spanning gaming, social, fintech, and cloud.
As an internal product, Hunyuan has no separate venture funding; it is built and operated by Tencent and served through Tencent Cloud, with open models distributed via Hugging Face and GitHub.
Hunyuan covers text generation (including the MoE-based HY3 with a 256K context window), image generation, 3D generation, and text-to-video and image-to-video generation up to 1080p. Several models are open-sourced, some under Apache 2.0.
Access is available three ways: the Yuanbao consumer app, the Tencent Cloud API on very low per-token pricing, and open-weight downloads for self-hosting.
Chinese consumers via Yuanbao, global developers and creators via open models, and enterprises via Tencent Cloud. The consumer experience is China-centric while the open models have broad international use.
Developers, creators, and researchers using open Hunyuan models for video and image generation, plus Chinese consumers using the Yuanbao app.
Engineering and platform teams selecting a low-cost LLM API or an open video/image model to self-host.
ML engineers, generative-media specialists, and cost-focused platform leads.
A developer team or creative studio wanting free or very low-cost, high-quality multimodal generation that can be self-hosted or accessed cheaply via API.
No separate funding disclosed for Hunyuan; it is an internal product of Tencent Holdings Ltd., a publicly traded company (HKEX: 0700).
Partly. Several Hunyuan models, including HunyuanVideo and the HY3 text model, have been open-sourced, some under Apache 2.0, while other tiers remain hosted-only via Tencent Cloud.
Yes. HunyuanVideo and its successors support text-to-video and image-to-video generation at up to 1080p, with cinematic camera motion and bilingual Chinese/English prompts.
It is among the cheapest available, with the HY3 model reported starting around $0.06 per million input tokens. Exact rates vary by provider and model; verify current pricing with Tencent Cloud.
Yuanbao is Tencent's consumer AI assistant app powered by Hunyuan models, offering chat and AI media features, primarily aimed at the Chinese market.
The open models are widely used globally, but the hosted consumer and cloud products are China-centric and carry data-residency and regulatory considerations for Western enterprises.
Side-by-side pages for pricing, features, and best-fit use cases.
Open-weight frontier LLM family from Z.ai (Zhipu AI), tuned for coding and agents.
Alibaba's free AI assistant, backed by the open-weight Qwen model family
Multimodal AI across text, video, audio, and music.
Baidu's flagship multimodal AI assistant for the Chinese market