Research
AI model API prices compared
What 72 AI models cost per million tokens on their standard API, from the reviews we keep current. The findings cover the 55 text and chat models billed for both input and output.
By ReviewRamble editors · Updated Nov 7, 2026 · How we built it
Key findings
- The dearest text model costs about 370 times the cheapest: GPT-6 Cyber at $40 per million blended tokens against Ling-3.0-flash-Sante at $0.108.
- The median closed model costs $3.88 per million blended tokens; the median open-weight model's hosted API costs $0.525 (24 closed and 31 open models).
- Output tokens typically cost 4 times input tokens (the median ratio), so long answers cost far more than long prompts.
- Of the tokens that the 14 priced models in OpenRouter's top 25 processed in the week to Nov 6, 2026, 59% went to models costing under $1 per million blended tokens.
Blended price per million tokens
All 72 models
Prices per million tokens in US dollars. Models in brackets are not text models and stay out of the findings. Usage is OpenRouter tokens in the latest week, for models in its top 25.
| Maker | Weights | Context | Cached input | OpenRouter tokens | ||||
|---|---|---|---|---|---|---|---|---|
| GLM-OCR(OCR & documents) | Z.ai (Zhipu AI) | Open weights | 131K tokens (config) | $0.03 | $0.03 | – | $0.03 | – |
| Ling-3.0-flash-Sante | Ant Group (inclusionAI) | Open weights | 256K tokens | $0.07 | $0.22 | – | $0.108 | – |
| LFM3 | Liquid AI | Open weights | 256K tokens (text model; 128K for vision and decision, 32K for audio and omni) | $0.08 | $0.2 | – | $0.11 | – |
| Laguna S 2.1 | Poolside | Open weights | 1M tokens | $0.1 | $0.2 | $0.01 | $0.125 | – |
| Phi-4-multimodal | Microsoft | Open weights | 128K tokens | $0.08 | $0.32 | – | $0.14 | – |
| Claude Haiku 5.5 | Anthropic | Closed weights | 1M tokens | $0.1 | $0.5 | $0.01 | $0.2 | 2.8T |
| GPT-6 Luna | OpenAI | Closed weights | 1.05M tokens | $0.1 | $0.5 | $0.01 | $0.2 | 7.3T |
| Qwen3.8-Flash-Next | Alibaba (Qwen) | Open weights | 262K tokens (1M extended and hosted) | $0.15 | $0.47 | $0.016 | $0.23 | – |
| GLM-5.3-Flash | Z.ai (Zhipu AI) | Open weights | 1M tokens | $0.15 | $0.5 | $0.03 | $0.237 | 15T |
| Nemotron 3.5 Super VL | NVIDIA | Open weights | 1M tokens | $0.15 | $0.5 | – | $0.237 | 3.4T |
| MiniMax-M3.1-Flash | MiniMax | Open weights | 1M tokens | $0.15 | $0.6 | $0.015 | $0.262 | 6T |
| Mistral Small 4 | Mistral AI | Open weights | 256K tokens | $0.15 | $0.6 | – | $0.262 | – |
| Ripple-65B Preview | Ricursive Intelligence | Open weights | 1M tokens | $0.2 | $0.6 | – | $0.3 | – |
| LFM2.5-24B-A2B | Liquid AI | Open weights | 131K tokens | $0.1 | $1 | – | $0.325 | – |
| MAI-Code-1.1-Flash | Microsoft AI | Closed weights | 256K tokens | $0.2 | $1.2 | $0.02 | $0.45 | – |
| Codestral 25.08 | Mistral AI | Closed weights | 128K tokens | $0.3 | $0.9 | – | $0.45 | – |
| Ling-3.1-flash | Ant Group (inclusionAI) | Open weights | 1M tokens | $0.3 | $0.9 | $0.06 | $0.45 | – |
| Moondream 3.1 (9B-A2B) | Moondream (M87 Labs) | Open weights | – | $0.3 | $1 | – | $0.475 | – |
| DeepSeek-V4.1-Flash | DeepSeek | Open weights | 1M tokens | $0.3 | $1.2 | – | $0.525 | 16.5T |
| LongCat-2.0 | Meituan | Open weights | 1M tokens | $0.3 | $1.2 | $0.006 | $0.525 | – |
| MiniMax-M3 | MiniMax | Open weights | 1M tokens | $0.3 | $1.2 | $0.06 | $0.525 | – |
| Muse Glimmer | Meta | Open weights | 131K tokens | $0.3 | $1.2 | – | $0.525 | – |
| MiMo-V2.6-Pro | Xiaomi | Open weights | 1M tokens | $0.435 | $0.87 | $0.004 | $0.544 | – |
| Command A+ | Cohere | Open weights | 128K tokens | $0.3 | $1.5 | – | $0.6 | – |
| Beam | Reflection | Open weights | 1M tokens | $0.4 | $1.2 | $0.04 | $0.6 | – |
| Gemini 3.5 Flash-Lite | Closed weights | 1M tokens | $0.3 | $2.5 | $0.03 | $0.85 | – | |
| Jina OCR v1(OCR & documents) | Jina AI (Elastic) | Open weights | 32K tokens | $0.5 | $2 | – | $0.875 | – |
| Mistral Large 4 | Mistral AI | Open weights | 1M tokens | $0.7 | $2.1 | $0.07 | $1.05 | – |
| Qwen3.8-27B | Alibaba (Qwen) | Open weights | 262K tokens (1M extended) | $0.5 | $3 | $0.1 | $1.13 | – |
| Hy4 preview | Tencent (Hy, formerly Hunyuan) | Open weights | 1M tokens | $0.834 | $2.5 | $0.042 | $1.25 | 6.6T |
| Step 5 Preview | StepFun | Open weights | 1M tokens | $1 | $2.7 | – | $1.43 | 6.7T |
| Gemini 3.8 Flash | Closed weights | 1M tokens | $0.75 | $3.75 | – | $1.5 | – | |
| Sakana Namazu | Sakana AI | Closed weights | 256K tokens (per trackers) | $0.95 | $4 | $0.15 | $1.71 | – |
| Inkling | Thinking Machines Lab | Open weights | 1M tokens | $1 | $4.05 | – | $1.76 | – |
| DeepSeek-V4-Pro | DeepSeek | Open weights | 1M tokens | $1.32 | $3.96 | – | $1.98 | – |
| GPT-6.1 Terra | OpenAI | Closed weights | 1.05M tokens | $1 | $5 | $0.05 | $2 | – |
| Muse Spark 1.3 | Meta | Closed weights | 1M tokens | $1.25 | $4.25 | $0.15 | $2 | – |
| Muse Spark Open | Meta | Open weights | 1M tokens | $1.25 | $4.25 | $0.15 | $2 | – |
| GLM-5.3 | Z.ai (Zhipu AI) | Open weights | 1M tokens | $1.4 | $4.4 | $0.26 | $2.15 | 3.6T |
| Mistral Medium 3.5 | Mistral AI | Open weights | 256K tokens | $1.5 | $7.5 | – | $3 | – |
| Grok 4.7 | SpaceXSI (formerly xAI) | Closed weights | 500K tokens | $2 | $6 | $0.5 | $3 | – |
| Qwen3.8-Max | Alibaba (Qwen) | Open weights | 1M tokens | $2 | $6 | – | $3 | – |
| MAI-Thinking-1 | Microsoft AI | Closed weights | 256K tokens | $2 | $8 | – | $3.5 | – |
| Aion 3.5 | Aion Labs (Deep Forge) | Closed weights | 262K tokens | $3 | $6 | $0.75 | $3.75 | – |
| Claude Sonnet 5.5 | Anthropic | Closed weights | 1M tokens | $2 | $10 | $0.1 | $4 | 2.3T |
| Gemini 4 Argon | Closed weights | 1M tokens | $2 | $10 | $0.1 | $4 | – | |
| Gemini 4 Carbon | Closed weights | 1M tokens | $2 | $10 | $0.1 | $4 | – | |
| GPT-6.1 Sol | OpenAI | Closed weights | 1.05M tokens | $2 | $10 | $0.1 | $4 | 3.6T |
| Kimi K3.1 Pro | Moonshot AI | Closed weights | 1M tokens | $2 | $10 | $0.2 | $4 | 3.5T |
| Gemini 3.5 Transcribe(Speech to text) | Closed weights | – | $2 | $12 | – | $4.5 | – | |
| Kimi K3 | Moonshot AI | Open weights | 1M tokens | $3 | $15 | $0.3 | $6 | 2.1T |
| Claude Opus 5.5 | Anthropic | Closed weights | 1M tokens | $4 | $20 | $0.2 | $8 | 6.5T |
| Muse Wonder | Meta | Closed weights | 1M tokens | $4 | $20 | $0.2 | $8 | – |
| GPT-Rosalind | OpenAI | Closed weights | Not stated | $5 | $25 | $0.5 | $10 | – |
| Claude Fable 5.1 | Anthropic | Closed weights | 1M tokens | $10 | $50 | $0.25 | $20 | – |
| GPT-6 Astra | OpenAI | Closed weights | 1.05M tokens | $10 | $50 | $1 | $20 | – |
| GPT-5.6-Cyber | OpenAI | Closed weights | 400K tokens | $12.5 | $75 | $1.25 | $28.13 | – |
| GPT-6 Cyber | OpenAI | Closed weights | 1,050,000 tokens | $20 | $100 | $2 | $40 | – |
| Perplexity Decider (pplx-decider)(Decision models) | Perplexity | Open weights | 262K tokens (trained to about 32K) | $0.02 | Free | – | – | – |
| DecisionNode(Decision models) | Bynn Intelligence | Closed weights | 64K tokens | $0.042 | Free | – | – | – |
| Jev(Decision models) | TypeSafe AI | Closed weights | 64K tokens (262K with long-context header) | $0.042 | Free | – | – | – |
| Microsoft-Decision-1(Decision models) | Microsoft AI | Closed weights | 32K tokens | $0.042 | Free | – | – | – |
| Jina Reranker v3.5(Embeddings & reranking) | Jina AI (Elastic) | Open weights | 131K tokens | $0.05 | – | – | – | – |
| ScaleDown Constrained Decoding Models (Classify, Extract, Summarize)(Decision models) | ScaleDown | Closed weights | 1M tokens | $0.05 | Free | – | – | – |
| Voyage rerank-3(Embeddings & reranking) | Voyage AI (MongoDB) | Closed weights | 32K tokens | $0.05 | – | – | – | – |
| voyage-4-large(Embeddings & reranking) | Voyage AI (MongoDB) | Closed weights | 32K tokens | $0.12 | – | – | – | – |
| text-embedding-3-large(Embeddings & reranking) | OpenAI | Closed weights | 8K tokens | $0.13 | – | – | – | – |
| Gemini Embedding 2(Embeddings & reranking) | Closed weights | 8K tokens | $0.2 | – | – | – | – | |
| Marengo 3.5(Embeddings & reranking) | TwelveLabs | Closed weights | – | $0.2 | – | – | – | – |
| Clef(Decision models) | Cloudflare | Open weights | 64K tokens | $0.24 | Free | – | – | – |
| Gemini 3.8 Flash TTS(Text to speech) | Closed weights | – | $0.5 | – | – | – | – | |
| Gemini Omni Flash 1.1(Video) | Closed weights | 1,048,576 tokens | $1.5 | – | – | – | – |
How we built it
- Prices. The standard list price per million text tokens from the maker's own API, or from the one host named in the review when the maker sells none. Batch, off-peak, long-context and promotional rates are left out. Each figure is copied from a review that records the date it was checked; we do not convert currencies, so models priced only in yuan or rupees are missing.
- Blended price. Three input tokens to one output token, (3 × input + output) ÷ 4, a common mix for chat and agent work. Your own mix may differ, so the table keeps input and output separate.
- Which models. Every model we review that publishes a clear per-token price. Models priced per character, second, image or minute (most speech, image and video models) are not here. The findings use text and chat models only.
- Usage. Tokens processed on OpenRouter in the week collected on Nov 6, 2026, from openrouter.ai/rankings. OpenRouter is one router among many, so this shows what its users run, not the whole market.
- Updates. The page is rebuilt whenever a review's price changes, and the usage figures refresh with our weekly count. We do not run the models ourselves for this page; quality and speed are in each review.
Using these figures
You are welcome to quote or chart these figures. Please credit ReviewRamble with a link to this page, and check the date, since prices change.
ReviewRamble, “AI model API prices compared”, Nov 7, 2026, https://www.reviewramble.com/research/ai-model-pricing