Openference catalog

Every model, one account

Sixteen frontier and open models behind one login — switch per session, per subagent, or per automation.

Don't pick a model. Pick a task.

Auto reads the job and routes every step — plans with the flagship, executes on the fast and cheap ones.

Modelctx
  • GLM-5.2Default256K
  • DeepSeek V4 Pro650K
  • MiniMax M31M
  • Kimi K2.7 Code256K
  • Qwen3.8 27B256K
  • Nemotron-3-120B256K
  • DeepSeek V4 Flash650K
  • Agnes 2.5 Flash512K
  • GLM-4.7 Flash128K

+ Kimi K2.6, Agnes Image 2.1 & 2.0 Flash, SDXL Lightning, SDXL Base 1.016 models in all

Routing through Openference means one account and one rate-limit pool across all sixteen models. You can also point Deyin at any OpenAI-compatible endpoint with your own keys.

Live pricing & stats on Openference

Per-model pricing, speed tiers, and usage numbers are maintained on Openference — Deyin routes through it, so numbers there are always current.