Hunyuan Hy3
Tencent’s own flagship, 295B MoE with 256K context — strong reasoning and multimodal depth.
Pay-as-you-go available
BD LLM Token Hub — DeepSeek, GLM, Kimi, Hunyuan & more through a single OpenAI-compatible key. Launching soon, supported locally by Malaysia’s Tencent Official Agentic Cloud Partner.
Switch between frontier models by changing one model name — no new accounts, no new bills. Pay-as-you-go rates are also available on top of your subscription.
Tencent’s own flagship, 295B MoE with 256K context — strong reasoning and multimodal depth.
Pay-as-you-go availableBest-value coding workhorse with up to 90% cache-hit savings — ideal for AI coding agents.
Pay-as-you-go availableAgent-orchestration specialist with 1M context — built for complex multi-step workflows.
Pay-as-you-go availableState-of-the-art long-context model with 1M context — deep research and document-scale reading.
Pay-as-you-go availableMultimodal frontier model — text, image and richer inputs in one economical API.
Pay-as-you-go availablePlus DeepSeek-V4-Pro, GLM 5.1 / 5.3, Kimi K2.6 / K3, MiniMax M3 / M2, MiMo, Hy-MT2 translation and Kinfra embeddings — new models added continuously.
GPT, Claude and Gemini are priced for their brands. The same agent tasks — coding, analysis, content — run on DeepSeek, GLM, Kimi and Hunyuan at a fraction of the cost. Same OpenAI-compatible API, same tools, far smaller bill.
Input $5.00 vs $0.15 · Output $30.00 vs $0.60 per 1M tokens — the same coding-agent workload, a fraction of the bill.
Input $5.00 vs $1.40 · Output $25.00 vs $4.40 per 1M tokens — 1M context agent orchestration without the Opus premium.
Input $2.00 vs $0.149 · Output $12.00 vs $0.596 per 1M tokens — Tencent’s own 295B MoE flagship at a workhorse price.
Input $2.00 vs $0.66 · Output $10.00 vs $1.98 per 1M tokens — flagship-grade reasoning at a mid-tier fraction.
Input $10.00 vs $3.00 · Output $50.00 vs $15.00 per 1M tokens — the same 1M context frontier class, for less than a third of the premium.
Last updated — vendor public list prices per 1M tokens, standard tier. Figures are vendor public list prices in USD per 1M tokens, standard tier — no batch, cache, promo or peak-hour discount applied. Vendors change list prices without notice: confirm on the vendor’s own pricing page before budgeting. Your savings start from an RM39/mo BD LLM Token Hub plan — request early access for launch pricing.
Spending on a US or global frontier API (OpenAI, Anthropic, Google)? Tell us your monthly bill — we estimate the same workload on China’s flagship models — DeepSeek, GLM, Kimi & Hunyuan — via BD LLM Token Hub. Live, in your currency.
China’s flagship models — DeepSeek, Zhipu GLM, Moonshot Kimi, Tencent Hunyuan — on one OpenAI-compatible API: you only change the base URL and key. Estimates use vendor public list prices (verified 19 Sep 2026) at your output ratio; actual savings vary with workload, caching and volume. Gemini 3.x Flash list price is introductory to 31 Dec 2026, then US$1.50 / US$7.50. Figures are vendor public list prices in USD per 1M tokens, standard tier — no batch, cache, promo or peak-hour discount applied. Vendors change list prices without notice: confirm on the vendor’s own pricing page before budgeting.
Pick a tier, get a pooled monthly quota of tokens, and switch between flagship models freely. Billed in MYR with a Malaysian invoice — no foreign-exchange surprises. Launching soon — request early access to be first in line.
Procurement-friendly plans for agencies, system integrators and enterprises that need shared access, per-key quotas and a Malaysian invoice.
50M+ tokens per month with raw-token billing, multiple keys and local onboarding.
Credit pools with advanced controls for larger deployments.
No lengthy sign-up flows. No juggling multiple dashboards. Request access now, then at launch grab your key and point your tools at it.
Tell us which plan you are interested in via the form on this page. We confirm your slot and notify you the moment we launch.
At launch, we deliver your key via ticket with the Singapore-region endpoint preconfigured — usually within the day.
Change the base URL and key in CodeBuddy Code, Cursor, Claude Code, OpenClaw, WorkBuddy & more. Done.
If your tool supports an OpenAI-compatible API, it works — just change the base URL and key. Here is what each one is, in plain English:
Powered by Tencent Cloud TokenHub — an OpenAI & Anthropic-compatible gateway, so existing clients only change base URL and key.
From customer-facing chatbots to autonomous coding agents — switch models per workload without switching vendors.
Multilingual support agents and FAQ bots on Kimi or GLM — with 24/7 reliability and MYR billing.
DeepSeek-V4.1-Flash and GLM-5.3 behind CodeBuddy Code, Cursor or Cline — at a fraction of direct vendor complexity.
Generate copy, scripts and campaign variants in volume — English, Bahasa Melayu or Chinese, from one API.
Long-context models like Kimi K3 power deep document Q&A and retrieval pipelines on your own knowledge base.
Everything Malaysian teams ask before switching to one LLM API key.
One subscription, one API key and one MYR bill — with access to DeepSeek, GLM, Kimi, Hunyuan and more. It is powered by Tencent Cloud TokenHub, a single OpenAI-compatible gateway, resold and supported locally by Big Domain.
Hunyuan Hy3, DeepSeek V4.1 Flash/Pro, GLM 5.1/5.3, Kimi K2.6/K3, MiniMax M2.7/M3, MiMo, Hy-MT2 and Kinfra embeddings — and you switch between them by changing one model name.
No. One TokenHub key covers every model in your plan. Sign up once, call any supported model through the same endpoint.
If your tool supports the OpenAI or Anthropic API (CodeBuddy Code, WorkBuddy, Cursor, Claude Code, Cline, OpenClaw and more), you only change the base URL and key — everything else stays the same.
Renew for the next month, or enable pay-as-you-go in the Tencent console for seamless overflow — you never get cut off mid-project.
Yes — Enterprise Lite and Enterprise Pro give you multiple API keys with per-key quotas and central management. Talk to Sales for a custom setup.
We are launching soon. Request early access and we will notify you first — early-access members get first API keys and launch pricing.
Complete the request form on this page and our team will confirm your slot, then keep you updated on the launch.
The frontier moves fast. These recent launches (China and US) are context for what you can build on — your plan and models on the hub above are what matter for pricing.
Updated 19 September 2026OpenAI’s most intelligent and most-aligned flagship yet — computer use, software engineering and security work at the frontier, and Astra Max now tops the CodeArena WebDev board (1,800 pts, 12 Sep 2026).
Agent-heavy professional tasks are its home turf.
EN · SiliconSnark ↗CN · 163.com · Zinc Industry ↗One stronger base model, two editions: Fable 5.1 for everyday work and Mythos 5.1 for security-first trusted access. Agent-heavy workloads run up to ~45% cheaper than Fable 5.
Caching costs cut too — cheaper long agent runs.
EN · SiliconSnark ↗CN · 163.com · Zinc Industry ↗Tencent’s open-weight flagship (Apache 2.0) — 770B total, 49B active, 1M context — with software-engineering results that rival Claude Opus 5 on Tencent’s Terminal-Bench 2.1 (85.4). Now served on Tencent Cloud TokenHub, the platform BD LLM Token Hub is built on.
Text-only preview — multimodal promised at production.
EN · Factlen ↗EN · Fello AI ↗CN · CSDN AI Daily (1 Sep) ↗Zhipu’s strongest open coding model to date — +50% over GLM-5.2 in internal evals, with cyber-defence capability rated on par with Mythos 5. One million-token context for complex agent workflows.
On the hub today.
EN · Alibaba ModelStudio ↗CN · CSDN AI Daily (4 Sep) ↗The 2.8T-parameter flagship with KDA hybrid attention and 1M context — deep research, document-scale reading and long-horizon agent tasks — now available in the Singapore region, where BD Token Hub is hosted.
Long-context specialist, live in SG.
EN · Reuters ↗CN · QQ News (31 Aug) ↗The flagship upgrade to the V4 line — 1.6T-parameter MoE (49B active), 1M context, with strong complex reasoning and professional code generation. The budget tier is now DeepSeek-V4.1-Flash (model id deepseek-flash) at off-peak US$0.15/M input and US$0.60/M output.
Both tiers are available on the hub today.
EN · DeepSeek API docs ↗EN · AI Pricing Guru ↗New to the lineup or unsure which model fits? Ask us ↗
Who makes the models — OpenAI, Anthropic, Google, the Chinese labs, NVIDIA and Malaysia’s own YTL AI Labs — and the flagship models each one ships today. Every brand page lists the model lineup, vendor list prices (USD ≈ RM) and the latest news with outlet links.
United States
China
Malaysia and open-enterprise
Full reference: /tokenhub/llm-brands/ — all 13 brand guides ↗
One partner for every digital layer – explore what else we build, host and grow for Malaysian businesses.
BD LLM Token Hub is coming soon. Tell us how you plan to use it and we will keep your spot warm — first API keys and launch pricing go to early-access members.