Model Tiers -- Hober Docs

Auto-Routing

Model Tiers

Hober organizes models into tiers based on capability and pricing. Use tiers to guide routing decisions and match models to your workload requirements.

FlagshipTop-tier reasoning and generation models. Best quality, higher cost.

deepseek/deepseek-reasoner-r1qwen/qwen3-235bkimi/kimi-k2

WorkhorseStrong general-purpose models. Great balance of quality and cost.

deepseek/deepseek-chat-v3.2deepseek/deepseek-coder-v3qwen/qwen3-coderglm/glm-4-pluskimi/moonshot-v1

EconomyLightweight and fast models. Lowest cost for high-throughput workloads.

qwen/qwen3-32bglm/glm-4-flashkimi/kimi-k2-mini

Tier-Based Routing

When using fallback chains, consider ordering models within the same tier for consistent quality, or across tiers for graceful degradation:

Same-Tier Fallback

"models": [
  "deepseek/deepseek-chat-v3.2",
  "qwen/qwen3-coder",
  "glm/glm-4-plus"
]

Cross-Tier Fallback

"models": [
  "deepseek/deepseek-reasoner-r1",
  "deepseek/deepseek-chat-v3.2",
  "glm/glm-4-flash"
]