Models & pricing
22 models in the catalog, 2 live now. Chat prices are USD per 1M tokens at the standard (cache-miss) rate; image models are priced per image. What you see is what you pay — no hidden multipliers. The lineup and prices update as upstream vendors change theirs.
China frontier models
DeepSeek V4 Flash
dabanzi/deepseek-v4-flashThe default workhorse — most-adopted model on OpenRouter. 1M context, thinking controllable.
Input $0.154 · Output $0.308 per 1M tokens · 1M context
DeepSeek V4 Pro
dabanzi/deepseek-v4-proDeepSeek flagship — frontier reasoning at a fraction of Western prices. 1M context.
Input $0.478 · Output $0.957 per 1M tokens · 1M context
Qwen3.7 Max
dabanzi/qwen3.7-maxAlibaba's flagship — highest LMArena rank of any Chinese lab. 1M context.
Input $2.75 · Output $8.25 per 1M tokens · 1M context
Qwen3.7 Plus
dabanzi/qwen3.7-plusBalanced Qwen tier — great price/performance for production.
Input $0.440 · Output $1.76 per 1M tokens · 1M context
Kimi K2.6
dabanzi/kimi-k2.6Moonshot flagship — multimodal, thinking modes, agentic tool use.
Input $1.04 · Output $4.40 per 1M tokens · 262K context
Kimi K2.7 Code
dabanzi/kimi-k2.7-codeKimi's strongest coding model — built for agentic coding.
Input $1.04 · Output $4.40 per 1M tokens · 262K context
Kimi K2.5
dabanzi/kimi-k2.5Best-value Kimi — multimodal input, agent tasks, context caching.
Input $0.660 · Output $3.30 per 1M tokens · 262K context
GLM-5.2
dabanzi/glm-5.2Z.ai flagship — top-scoring Chinese model on Artificial Analysis. 1M context.
Input $1.54 · Output $4.84 per 1M tokens · 1M context
GLM-4.7 FlashX
dabanzi/glm-4.7-flashxCheapest fast tier in the catalog — bulk and agent loops.
Input $0.077 · Output $0.440 per 1M tokens · 200K context
MiniMax M3
dabanzi/minimax-m3MiniMax flagship — natively multimodal (image + video input), 1M context, agent-grade.
Input $0.330 · Output $1.32 per 1M tokens · 1M context
MiMo V2.5 Pro
dabanzi/mimo-v2.5-proXiaomi's 1T-param flagship — top-5 OpenRouter adoption, MIT-licensed weights.
Input $0.478 · Output $0.957 per 1M tokens · 1M context
MiMo V2.5
dabanzi/mimo-v2.5Omnimodal MiMo — text, image, video and audio input at a bargain price.
Input $0.154 · Output $0.308 per 1M tokens · 1M context
Qwen3 Coder
qwen/qwen3-coderQwen3 Coder — agentic coding, very cheap.
Input $0.242 · Output $1.98 per 1M tokens · 256K context
Western & open frontier
GPT-5
openai/gpt-5OpenAI flagship — best general reasoning + tool use.
Input $1.38 · Output $11.00 per 1M tokens · 400K context
GPT-5 Mini
openai/gpt-5-miniCheap, fast OpenAI tier for bulk + agents.
Input $0.275 · Output $2.20 per 1M tokens · 400K context
GPT-4o
openai/gpt-4oOpenAI multimodal workhorse.
Input $2.75 · Output $11.00 per 1M tokens · 128K context
Grok 4
xai/grok-4xAI Grok 4 — strong reasoning.
Input $3.30 · Output $16.50 per 1M tokens · 256K context
Mistral Large
mistralai/mistral-largeMistral's flagship — multilingual + code.
Input $2.20 · Output $6.60 per 1M tokens · 128K context
Llama 3.3 70B
meta-llama/llama-3.3-70bMeta Llama 3.3 70B, served fast on Groq.
Input $0.649 · Output $0.869 per 1M tokens · 128K context
Image models
Seedream 4.5
dabanzi/seedream-proByteDance Seedream 4.5. Best-in-class CJK text rendering, 4K.
standard: $0.033/image · 4k: $0.044/image
Seedream 5.0
dabanzi/seedream-ultraSeedream 5.0 flagship — character consistency + web retrieval.
standard: $0.050/image · 4k: $0.077/image
Seedream 5.0 Lite
dabanzi/seedream-liteLite tier — best value, billed per image.
standard: $0.024/image
“Coming soon” models are in the catalog but their upstream connection isn't enabled yet — they don't appear in GET /v1/modelsand can't be called or billed until they go live.