Compare AI API costs across providers
Find a model for what you’re building. Explore the latest pricing and benchmarks, or ask for a comparison tailored to your workload.
Tell us what you’re building. Compare suggested models, estimated costs and available benchmarks.
62 reviewed offers · 11 providers · price review: Friday 16th October 2026
Model news
Recent stories about AI models, from public sources.
Meta AI / Meta Developer Blog
2026-10-15
Meta published Muse Spark Open, an open-weight 816B-parameter mixture-of-experts model that activates ~42B parameters per token, accepts text, image, video and audio input, and supports a 1 million‑token context window; Meta says the weights and runtime support are available for deployment, including optimizations for Mac Studio.
Read the source ↗NVIDIA Developer / NVIDIA Blog
2026-10-15
NVIDIA announced Nemotron 3.5 Super VL, an open 120B-parameter multimodal mixture-of-experts model (with a smaller Bolt 74B variant) positioned for specialized always-on agents and multimodal reasoning, and published technical and deployment details for developers.
Read the source ↗Microsoft AI (Lab Notes)
2026-10-15
Microsoft AI posted a Lab Notes entry saying it completed pre-training MAI-Base-2, a mixture-of-experts foundation model with 2.8 trillion total parameters and 80 billion active parameters, and described next steps (post-training) and comparisons with open-weight models.
Read the source ↗Find your starting point
What matters most to you? Switch groups to compare up to five models on the same axes.
Highest general Intelligence Index scores among models with both chart measurements. This is not a task-specific ranking. 167 of 686 benchmark variants have both chart measurements. Scores are Intelligence Index v4.3, not percentages. Missing measurements are excluded; this is not the entire API market.
Higher intelligence ↑ · Lower benchmark cost ←
Axes stay fixed between groups. Numbers match the models alongside the chart. One benchmark variant per model name. Cost/task is a benchmark measurement, not a monthly estimate.
Source: Artificial Analysis · snapshot 2026-09-30 · General scores do not guarantee results on your task. Explore the full benchmark dataset →
Start with the workload you recognise. These guides explain the assumptions before you compare costs.
Already considering a provider? Compare its rates, explore alternatives or read a workload guide.
Sources
Pricing uses dated provider snapshots and validation sources. Intelligence and performance use Artificial Analysis benchmarks. Missing evidence stays marked as unavailable; token estimates do not include every service fee.
Coverage
Compare reviewed API offers on the same workload. Coverage is incomplete; unreviewed legacy rows remain in the downloadable audit data. Primary-source review dated Friday 16th October 2026. Use the calculator above for monthly estimates; browse the reference rates below for details.
Structured dataset: ai-provider-pricing-validated.json · Methodology: AI API pricing validation pipeline · Long-form article
Reference rates
Official-style input, cached-input, and output rates used by the calculator. Monthly spend depends on your message volume, average tokens per message, and cache hit rate — not headline list prices alone.
| Model | Provider | Input / MTok | Cached / MTok | Output / MTok |
|---|---|---|---|---|
| Claude Fable 5 (standard) | Anthropic | $10.00 | $1.00 | $50.00 |
| Claude Fable 5.1 (standard) | Anthropic | $10.00 | $0.25 | $50.00 |
| Claude Haiku 4.5 (standard) | Anthropic | $1.00 | $0.10 | $5.00 |
| Claude Opus 4.5 (standard) | Anthropic | $5.00 | $0.50 | $25.00 |
| Claude Opus 4.6 (standard) | Anthropic | $5.00 | $0.50 | $25.00 |
| Claude Opus 4.7 (standard) | Anthropic | $5.00 | $0.50 | $25.00 |
| Claude Opus 5 (standard) | Anthropic | $5.00 | $0.50 | $25.00 |
| Claude Opus 5.5 (standard) | Anthropic | $4.00 | $0.20 | $20.00 |
| Claude Sonnet 4.5 (standard) | Anthropic | $3.00 | $0.30 | $15.00 |
| Claude Sonnet 4.6 (standard) | Anthropic | $3.00 | $0.30 | $15.00 |
| Claude Sonnet 5 (standard) | Anthropic | $2.00 | $0.20 | $10.00 |
| Amazon Nova Lite (standard) | AWS Bedrock | $0.06 | $0.01 | $0.24 |
| Amazon Nova Micro (standard) | AWS Bedrock | $0.04 | $0.01 | $0.14 |
| Amazon Nova Pro (Latency Optimized) (latency optimized) | AWS Bedrock | $1.00 | — | $4.00 |
| Amazon Nova Pro (standard) | AWS Bedrock | $0.80 | $0.20 | $3.20 |
| Nova 2 Lite (standard) | AWS Bedrock | $0.33 | $0.08 | $2.75 |
| GPT-5 Codex Global (standard) | Azure OpenAI | $1.25 | $0.13 | $10.00 |
| GPT-5-mini Global (standard) | Azure OpenAI | $0.25 | $0.03 | $2.00 |
| Command R (standard) | Cohere | $0.15 | — | $0.60 |
| Command R+ (standard) | Cohere | $2.50 | — | $10.00 |
| DeepSeek V4-1-FLASH (off-peak) | DeepSeek | $0.15 | $0.00 | $0.60 |
| DeepSeek V4-1-FLASH (peak) | DeepSeek | $0.30 | $0.01 | $1.20 |
| DeepSeek V4-PRO-0813 (off-peak) | DeepSeek | $0.66 | $0.02 | $1.98 |
| DeepSeek V4-PRO-0813 (peak) | DeepSeek | $1.32 | $0.04 | $3.96 |
| Gemini 2.5 Flash (standard) | Google Gemini | $0.30 | $0.03 | $2.50 |
| Gemini 2.5 Flash-Lite (batch) | Google Gemini | $0.05 | $0.01 | $0.20 |
| Gemini 2.5 Flash-Lite (standard) | Google Gemini | $0.10 | $0.01 | $0.40 |
| Gemini 2.5 Pro (batch) | Google Gemini | $0.63 | $0.13 | $5.00 |
| Gemini 2.5 Pro (standard) | Google Gemini | $1.25 | $0.13 | $10.00 |
| Gemini 3.8 Flash (standard) | Google Gemini | $0.75 | $0.07 | $3.75 |
| MiniMax M2.5 (standard) | MiniMax | $0.30 | $0.03 | $1.20 |
| MiniMax M2.7 (standard) | MiniMax | $0.30 | $0.06 | $1.20 |
| MiniMax M3 (standard) | MiniMax | $0.30 | $0.06 | $1.20 |
| Ministral 3 14B (standard) | Mistral | $0.20 | $0.02 | $0.20 |
| Ministral 3 3B (standard) | Mistral | $0.10 | $0.01 | $0.10 |
| Mistral Large 3 (standard) | Mistral | $0.50 | $0.05 | $1.50 |
| GPT-5.4 (standard) | OpenAI | $2.50 | $0.25 | $15.00 |
| GPT-5.4 mini (standard) | OpenAI | $0.75 | $0.07 | $4.50 |
| GPT-5.4 nano (standard) | OpenAI | $0.20 | $0.02 | $1.25 |
| GPT-5.5 (standard) | OpenAI | $5.00 | $0.50 | $30.00 |
| GPT-5.5 Pro (standard) | OpenAI | $30.00 | — | $180.00 |
| GPT-5.6 Luna (standard) | OpenAI | $0.20 | $0.02 | $1.20 |
| GPT-5.6 Sol (standard) | OpenAI | $4.00 | $0.40 | $20.00 |
| GPT-5.6 Terra (standard) | OpenAI | $2.00 | $0.20 | $12.00 |
| GPT-6 Astra (standard) | OpenAI | $10.00 | $1.00 | $50.00 |
| GPT-6 Luna (standard) | OpenAI | $0.10 | $0.01 | $0.50 |
| GPT-6 Sol (standard) | OpenAI | $2.00 | $0.20 | $10.00 |
| Gemini 2.5 Flash (flex batch) | Vertex AI | $0.15 | — | $1.25 |
| Gemini 2.5 Flash (priority) | Vertex AI | $0.54 | $0.05 | $4.50 |
| Gemini 2.5 Flash (standard) | Vertex AI | $0.30 | $0.03 | $2.50 |
| Gemini 2.5 Flash Lite (priority) | Vertex AI | $0.18 | $0.02 | $0.72 |
| Gemini 2.5 Flash Lite (standard) | Vertex AI | $0.10 | $0.01 | $0.40 |
| Gemini 2.5 Pro (flex batch) | Vertex AI | $0.63 | — | $5.00 |
| Gemini 2.5 Pro (priority) | Vertex AI | $2.25 | $0.23 | $18.00 |
| Gemini 2.5 Pro (standard) | Vertex AI | $1.25 | $0.13 | $10.00 |
| Grok 4.20 Multi-Agent (standard) | xAI | $1.25 | $0.20 | $2.50 |
| Grok 4.20 Non-Reasoning (standard) | xAI | $1.25 | $0.20 | $2.50 |
| Grok 4.20 Reasoning (standard) | xAI | $1.25 | $0.20 | $2.50 |
| Grok 4.3 (standard) | xAI | $1.25 | $0.20 | $2.50 |
| Grok 4.5 (standard) | xAI | $2.00 | $0.30 | $6.00 |
| Grok 4.6 (standard) | xAI | $2.00 | $0.50 | $6.00 |
| Grok 4.7 (standard) | xAI | $2.00 | $0.50 | $6.00 |
How it works
Need more detail on presets, token mix, and reading the ranking table? Read the full how-to guide →
Pick a preset (support bot, code assistant, RAG, or agent) or enter requests per month and average tokens per request.
Scan the top recommendations for this workload — ranked by monthly token cost, with source conflicts filtered out by default.
Select a recommendation to inspect the monthly estimate, assumptions, and input/cached/output breakdown.
Open Advanced token mix for cache-aware splits, or expand View all models to search and filter the full ranked list.
FAQ
need help choosing models?
Scale AI products without billing surprises — from experimentation to production.
