Models
The Coding Plan supports a variety of mainstream AI models, with different subscription tiers having access to different models and routing strategies. The system dynamically adjusts routing based on network conditions, load, and your subscription usage stage.
The table below lists all currently supported model IDs and their routing strategies under different subscription plans.
| Model ID (API Model) | Display Name | Go | Pro | Max / Ultra |
|---|---|---|---|---|
gemini-3.5-flash | Gemini 3.5 Flash | Anti First | Anti First | Hybrid (GCP/Anti) |
gemini-3.1-pro-preview | Gemini 3 Pro Preview | Anti First | Anti First | Hybrid (GCP/Anti) |
gemini-3.1-flash-lite | Gemini 3.1 Flash Lite | Anti First | Anti First | Hybrid (GCP/Anti) |
gemini-3-flash-preview | Gemini 3 Flash Preview | Anti First | Anti First | Hybrid (GCP/Anti) |
claude-opus-4-8 | Claude Opus 4.8 | Degraded | Kiro First ⚠️ | Hybrid |
claude-opus-4-7 | Claude Opus 4.7 | Degraded | Kiro First ⚠️ | Hybrid |
claude-opus-4-6 | Claude Opus 4.6 | Degraded | Kiro First ⚠️ | Hybrid |
claude-sonnet-4-6 | Claude Sonnet 4.6 | Degraded | Kiro First ⚠️ | Hybrid |
claude-opus-4-5 | Claude Opus 4.5 | Degraded | Kiro First ⚠️ | Hybrid |
claude-sonnet-4-5 | Claude Sonnet 4.5 | Degraded | Kiro First ⚠️ | Hybrid |
claude-haiku-4-5 | Claude Haiku 4.5 | Degraded | Kiro First ⚠️ | Hybrid |
gpt-5.5 | GPT-5.5 | Codex Routing | Codex Routing | Codex Routing |
gpt-5.4 | GPT-5.4 | Codex Routing | Codex Routing | Codex Routing |
gpt-5.4-mini | GPT-5.4 Mini | Codex Routing | Codex Routing | Codex Routing |
gpt-5.3-codex | GPT-5.3 Codex | Codex Routing | Codex Routing | Codex Routing |
grok-4.5 | Grok 4.5 | Grok Routing | Grok Routing | Grok Routing |
grok-4.3 | Grok 4.3 | Grok Routing | Grok Routing | Grok Routing |
composer-2.5 | Composer 2.5 | Cursor Routing | Cursor Routing | Cursor Routing |
deepseek-v4-flash | DeepSeek V4 Flash | Deepseek Routing | Deepseek Routing | Deepseek Routing |
deepseek-v4-pro | DeepSeek V4 Pro | Deepseek Routing | Deepseek Routing | Deepseek Routing |
GLM-5.1 | GLM-5.1 | Direct | Direct | Direct |
GLM-4.7 | GLM-4.7 | Volcano Engine | Direct/Volcano | Direct/Volcano |
GLM-4.6 | GLM-4.6 | Volcano Engine | Direct/Volcano | Direct/Volcano |
GLM-4.5-Air | GLM-4.5 Air | Direct | Direct | Direct |
MiniMax-M3 | MiniMax M3 | Direct | Direct | Direct |
MiniMax-M2.7 | MiniMax M2.7 | Direct | Direct | Direct |
MiniMax-M2.5 | MiniMax M2.5 | Direct | Direct | Direct |
kimi-k2.7-code | Kimi k2.7 Code | Volcano Engine | Hybrid (Direct/Volc) | Hybrid (Direct/Volc) |
kimi-k2.6 | Kimi k2.6 | Volcano Engine | Hybrid (Direct/Volc) | Hybrid (Direct/Volc) |
Routing Strategy Description:
- CC Routing (claude-code): Official high-availability channel, highest stability, higher price.
- Codex Routing: OpenAI-specific code optimization channel, optimized for programming context.
- Hybrid Routing: Automatically switches between GCP, Azure, Antigravity (Anti), and official sources based on load, ensuring high availability.
- Kiro / Anti / Azure: Premium enterprise-grade distribution channels, cost-effective.
- ⚠️ Dynamic Degradation: Indicates that when "economic mode" is enabled or "quota exceeded", the model has a higher probability of being degraded (e.g., switched to MiniMax, GLM, or smaller parameter models) to maintain service availability.
Regarding Quota and Degradation: When your subscription usage exceeds 90%, the system will forcefully enable load balancing strategy. Max/Pro users after exceeding quota may not be able to access Gemini or Claude through priority channels like GCP/CC, and will fall back to standard channels or more cost-effective models (such as MiniMax/Azure).