Models
One balance. Every model that matters.
Credits spend the same on any model — the only difference is how many requests they buy. Estimates per 1,000 credits, one typical coding request: ~10k input, ~1k output tokens.
GLM-5.3
Z.aiDefault
≈ 1,086
requests / 1,000 credits
DeepSeek V4
DeepSeek
≈ 19,263
requests / 1,000 credits
GPT-6 Astra
OpenAI
≈ 133
requests / 1,000 credits
Claude Sonnet 5
Anthropic
≈ 666
requests / 1,000 credits
GPT-5.6 Sol
OpenAI
≈ 250
requests / 1,000 credits
≈ 60 credits runs every model on this page once. A Starter month does that 16× over.
Get 1,000 creditsPopular
The models people actually reach for.
- GLM-5.3Z.ai≈ 1,086
- Kimi K2.7 CodeMoonshot AI≈ 2,000
- Kimi K3Moonshot AI≈ 990
- Claude Sonnet 5Anthropic≈ 666
- GPT-5.3 CodexOpenAI≈ 888
- GPT-5.6 TerraOpenAI≈ 625
- Gemini 3 ProGoogle≈ 625
- Gemini 3.5 FlashGoogle≈ 833
- Grok Code Fast 2xAI≈ 5,000
- Qwen3.7 PlusAlibaba≈ 4,464
- Kimi K2.7Moonshot AI≈ 2,352
- DeepSeek R4DeepSeek≈ 4,347
Frontier
The heavyweights, for tasks that need them.
- GPT-6 AstraOpenAI≈ 133
- Claude Opus 5Anthropic≈ 266
- Claude Fable 5Anthropic≈ 133
- GPT-5.6 SolOpenAI≈ 250
- Grok 4.6xAI≈ 769
- Qwen3.7 MaxAlibaba≈ 1,230
- Kimi K2.7 ThinkingMoonshot AI≈ 2,352
- Mistral Large 3Mistral AI≈ 769
Value
The workhorses that stretch a credit furthest.
- GLM-5.3 FlashZ.ai≈ 20,000
- Qwen3 Coder NextAlibaba≈ 10,000
- DeepSeek V4DeepSeek≈ 19,263
- Devstral 2Mistral AI≈ 3,030
- Claude Haiku 4.5Anthropic≈ 1,333
- GPT-5.6 LunaOpenAI≈ 6,250
- Gemini 3 FlashGoogle≈ 2,500
- Grok 4 FastxAI≈ 8,000
- MiniMax M3MiniMax≈ 4,761
- Llama 4 MaverickMeta≈ 9,523
- Mistral Medium 3.5Mistral AI≈ 2,985
- Command A+Cohere≈ 3,571
Estimates, not quotes: actual usage varies with context size, cache hits, and live provider rates — metered usage is exact. The plans run the same models — they differ in monthly credits and caps, never in model access.