GLM-5.3Now available
Try it now
Limited-time 50% offDouble 11 sale

GLM 5.3 limited-time offer. Save up to 50%.

More power for reasoning, coding and agent workflows, now at a special price.

Input¥4/ 1M tokens
Output¥14/ 1M tokens
Recharge now

  • GLM 5.3 special offer

    Save up to 50%

    For a limited time

  • The same rate, all day

    Day or night, with no peak hours to avoid

  • Lower cache-hit prices

    Cached input ¥1 / 1M tokens

  • One million tokens

    Long documents, larger codebases, lasting context

Model marketplace

Explore leading open-source and proprietary models. Deploy and start calling in minutes.

View all models

MiniMax H3

Open-weight omni-modal video model with joint audio-video output, supporting text-to-video, first/last-frame and omni-reference generation.

Try now
Product interface preview
View all models

TokenFab earns CAICT certificationToken Service Capability Climbing Program 2.0

One of the first three organizations nationwide

  • TPS ≥65/s
  • TTFT ≤0.9s
  • Success rate ≥99.9%

In October 2026, TokenFab passed the general-scenario baseline of CAICT’s Token Service Capability Climbing Program 2.0 with DeepSeek and GLM models. View details

Easy accessYour first model

Sign up for free credits and make your first call in minutes.

Access leading open-source and proprietary models through one API.

Get API Key

No text model is currently available. See the docs to learn how to connect.

Read docs ↗
MiniMax H3

Coming soon

Compute services

Elastic, reliable and secure compute, from shared inference to dedicated clusters.

Dedicated clusters

Physically isolated, exclusive compute resources for security, compliance and consistent performance.

Physical isolationExclusive compute
Learn more

Compute hosting

Host GPU servers and clusters with 120kW+ liquid-cooled racks, redundant power and 24/7 on-site operations to bring AI assets into production faster.

High-grade facilitiesFull visibility and controlLayered protectionReliable and trusted
Learn more
MLPS Level 3Physical data isolation99.99% SLA24/7 support

Coming soon

Agent Applications

Understand goals, use professional tools, and deliver documents, analysis, and applications. Turn ideas into results.

Coming soon

Model production

One workbench for distributed training, no-code fine-tuning, evaluation and deployment.

Model production workflow: business data, fine-tuning, evaluation, deployment and applications for enterprise use, XY Work and XY Family.

Sign up and start your AI journey

Register now