TokenMix Research Lab · 2026-06-15

Last Updated: 2026-06-15 Author: TokenMix Research Lab Data verified: 2026-06-15 — Tencent Cloud Hunyuan pricing, OpenRouter model listing (tencent/hy3-preview), Artificial Analysis Intelligence Index, TokenMix.ai model tracker

Tencent Hunyuan API Pricing 2026: HY3 & HY2.0 English Access

Tencent's Hunyuan models are among the cheapest capable LLMs in the world — HY3 Preview runs about $0.063 per 1M input / $0.21 per 1M output on OpenRouter — but actually accessing them from outside China is the real problem. Tencent Cloud's primary console assumes a mainland account, some Hunyuan products have been geo-blocked in the EU/UK/South Korea, and the in-house HY2.0 line took a steep price hike in March 2026. This guide gives you the verified prices for the whole Hunyuan line and the practical English-access paths.

HY3 Preview launched April 23, 2026 as an open-weight model under the Tencent Hunyuan Community License — a 295B-total / 21B-active Mixture-of-Experts design with a 256K context window, text-only. The older HY2.0 Instruct and Think models remain in service but got dramatically more expensive this spring. We pulled every live number and mapped the access routes below.

Table of Contents

• Quick Verdict
• Quick Comparison: The Hunyuan Line
• HY3 Preview Pricing
• The March 2026 HY2.0 Price Hike
• Benchmarks: Claimed and Verified
• Cost per Task: Real Math
• The Real Problem: English / International Access
• How to Access Hunyuan from Outside China
• Decision Guide
• Final Recommendation
• FAQ
• Related Articles

Quick Verdict

The 40-60 word takeaway: Hunyuan HY3 Preview is one of the cheapest capable open-weight models in 2026 (~$0.063/$0.21), with an independent intelligence score of 42. The hard part is access: Tencent Cloud is China-account-centric and partly geo-blocked, so international developers should reach Hunyuan through OpenRouter or a unified gateway, not the native console.

Quick Comparison: The Hunyuan Line

HY2.0 USD figures are approximate conversions from Tencent Cloud's post-hike CNY list at ~6.89 CNY/USD. HY3 Preview pricing per OpenRouter's tencent/hy3-preview listing. The striking thing: the newest, open-weight model (HY3) is roughly 10x cheaper than the older closed HY2.0 line after the hike.

HY3 Preview Pricing

HY3 Preview is the model to care about, and it is remarkably cheap. Based on OpenRouter's listing, the verified rate is $0.063 per 1M input / $0.21 per 1M output. For context, that input price is roughly 1/80th of GPT-5.5's and undercuts even DeepSeek on raw token cost.

Because HY3 is open-weight, you have routing choices, and the per-token rate genuinely varies by provider:

We will be straight about this, because the whole point of TokenMix.ai is neutrality: for raw HY3 Preview tokens, OpenRouter's community route is cheaper than our gateway. Where a unified gateway earns its rate is when you also need the closed HY2.0 models (not on OpenRouter), want one OpenAI-compatible key across 300+ models with fallback routing, or want the neutral price tracking that flagged the HY2.0 hike in the first place. If all you need is the cheapest HY3 tokens and nothing else, route it through OpenRouter or one of its alternatives — that is the honest answer.

The March 2026 HY2.0 Price Hike

If you were running HY2.0 in production, you felt this. Effective 2026-03-13, Tencent raised HY2.0 prices steeply:

A 4–5x overnight increase is the kind of event that should make any team reconsider single-vendor dependence. It is also exactly why neutral price tracking matters: TokenMix.ai's model tracker logged the change, and on the Hunyuan model pages you can see current versus historical rates rather than discovering a hike on your next invoice.

There is a practical upshot. Through the TokenMix.ai gateway, HY2.0 Instruct lists at $0.422/$1.06 and HY2.0 Think at $0.528/$2.11 — roughly 30–35% below the post-hike Tencent Cloud direct rates. For the closed HY2.0 models specifically, the gateway is cheaper than going direct, and it sidesteps the China-account requirement. So the routing answer flips by model: cheapest HY3 is OpenRouter; cheapest HY2.0 is the gateway.

Benchmarks: Claimed and Verified

HY3 Preview has one independent data point and several vendor claims.

Independently verified:

Self-reported by Tencent (treat as claims):

Read this honestly: an Artificial Analysis score of 42 places HY3 in the mid-pack, not at the frontier. It is well behind Qwen 3.7 Max (56.6) and the top closed models. The self-reported coding number (74.4) looks strong, but it sits awkwardly above what a 42 intelligence index would predict, so apply a discount until it is independently reproduced.

The right framing for HY3 is value, not peak capability: it is a cheap, competent open-weight model for high-volume, cost-sensitive tasks — not the model you reach for when you need the smartest possible answer. Compare it against the field on the best Chinese AI models guide.

Cost per Task: Real Math

Take a standard task: 100K input, 20K output.

Two clear conclusions: HY3 Preview is almost free at this scale — about one cent per task on OpenRouter. And for the HY2.0 line, the gateway route is roughly 32% cheaper than Tencent Cloud direct after the March hike.

Scale HY3 Preview to 1B input / 200M output per month (a serious high-volume pipeline):

• OpenRouter: $63 + $42 = $105/month
• TokenMix: $163 + $109 = $272/month

For a billion input tokens, you are paying in the low hundreds either way. HY3's reason to exist is exactly this: bulk, cost-sensitive text work where mid-pack quality is good enough. For modeling this across providers, see our pricing calculator guide.

The Real Problem: English / International Access

For most international developers, Hunyuan's blocker is not price — it is access. Three frictions to plan around:

• China-account-centric console. Tencent Cloud's primary Hunyuan console and documentation are oriented to mainland accounts. The international Tencent Cloud portal exists, but Hunyuan onboarding, billing, and docs are less complete in English than the Chinese-language experience.
• Geo-blocking precedent. Some Hunyuan products have been restricted in the EU, UK, and South Korea. Availability varies by product and region, so do not assume the native API is reachable from your jurisdiction without checking.
• Documentation gaps. English API docs for the Hunyuan line lag the Chinese versions, which slows integration and debugging.

This is the core reason TokenMix.ai's Chinese-model coverage exists: surfacing models like Hunyuan in English, with verified pricing and an OpenAI-compatible interface, so a developer in Berlin or São Paulo can evaluate and call them without a mainland-China account.

How to Access Hunyuan from Outside China

For most developers outside China, the practical answer is: HY3 Preview → OpenRouter for cheapest tokens; the full Hunyuan line (including HY2.0) → a unified gateway. Both expose the OpenAI-compatible format, so the integration is a base-URL and model-name change. The OpenAI-compatible API gateway guide has the SDK pattern.

Decision Guide

Final Recommendation

Based on the data as of 2026-06-15: Tencent Hunyuan HY3 Preview is a genuine bargain for high-volume, cost-sensitive text work — about a cent per typical task — but it is mid-pack on quality and a hassle to reach natively. Treat it as a bulk-throughput option, not a frontier model. Route HY3 through OpenRouter for the cheapest tokens; route the closed HY2.0 line through a unified gateway, where it runs ~32% below Tencent Cloud's post-hike direct pricing and skips the mainland-account requirement.

If you need top-tier reasoning, look at Qwen 3.7 Max or Claude Opus instead. If you need cheap, competent, English-accessible Chinese-model tokens at scale, Hunyuan earns a spot in your routing table. You can compare its live pricing and independent scores against 300+ models on TokenMix.ai before wiring it in.

FAQ

How much does the Tencent Hunyuan API cost?

HY3 Preview is about $0.063 per 1M input and $0.21 per 1M output via OpenRouter. The older HY2.0 line, after a March 2026 hike, runs roughly $0.65/$1.62 (Instruct) and $0.77/$3.08 (Think) on Tencent Cloud direct, or about 32% less through a gateway.

Why did Hunyuan HY2.0 prices go up?

Tencent raised HY2.0 list prices effective 2026-03-13 — roughly +460% on Instruct and +430% on Think. HY3 Preview, released the following month as an open-weight model, is far cheaper and is the better value going forward.

Can I use Hunyuan from outside China?

Yes, but not easily through Tencent Cloud's native console, which is mainland-account-centric and partly geo-blocked (EU/UK/South Korea). The practical routes for international developers are OpenRouter (for open-weight HY3) or a unified gateway like TokenMix.ai (for the full Hunyuan line).

Does Hunyuan HY3 have open weights?

Yes. HY3 Preview ships under the Tencent Hunyuan Community License (295B total / 21B active MoE, 256K context, text-only), so you can self-host it or rent it from third-party providers. The HY2.0 line is closed.

How good is Hunyuan HY3 compared to Qwen or DeepSeek?

Mid-pack. HY3 scores 42 on the independent Artificial Analysis Intelligence Index, behind Qwen 3.7 Max (56.6). Its strength is price, not peak capability — it is a bulk, cost-sensitive option rather than a frontier model.

What is the cheapest way to use Hunyuan?

For HY3 Preview, OpenRouter's community route at $0.063/$0.21 is the cheapest per token. For the closed HY2.0 models, a unified gateway is cheapest, running roughly 32% below Tencent Cloud's post-hike direct pricing.

Is Hunyuan HY3 multimodal?

No. The HY3 Preview text model is text-only. Tencent has separate vision and video models in the Hunyuan family, but HY3 Preview itself does not handle images.

Related Articles

• Hunyuan-TurboS Review: Tencent's Hybrid Mamba-Transformer MoE (2026)
• Hunyuan-T1 Review: Tencent's Deep-Reasoning Rival to DeepSeek R1 (2026)
• Best Chinese AI Models 2026: Kimi, DeepSeek, Qwen, GLM Compared
• Best OpenRouter Alternatives 2026: 8 API Options Compared
• AI API Gateway 2026: 7 LLM Routing and Fallback Options

About TokenMix

TokenMix.ai is a neutral AI model intelligence platform. We independently track pricing, benchmarks, and API reliability for 300+ models, and provide a single OpenAI-compatible gateway to access them — often well below official rates. We don't represent any model vendor; our job is to tell developers what's actually true when they pick a model or an API. Check live prices on /models, compare plans on /pricing, or read the integration docs at /docs.