The AI Loadout

AI model API price comparison

Checked on

For most apps, a mid-priced API such as Claude Sonnet 5.5 ($2.00 in, $10.00 out per million tokens) is the default. Cheap, high-volume calls can go to GPT-6 Luna ($0.10 in, $0.50 out). Reach for GPT-6 Astra ($10.00 in, $50.00 out) only for the hardest reasoning. Every figure is the vendor price on the linked source page, checked Nov 14, 2026. Intro and launch-sale prices are shown as the vendor listed them.

Building an app on these APIs? See the best AI model API setup. Looking for the interactive cost calculator and historical change log? See the AI Price Index. Running on your own machine instead? See best local models for Ollama and LM Studio.

Price per million tokens

38 active hosted models with an official source URL and real prices. Sorted by the first dollar amount in the input-price string. Local models are on the local list, not here.

ModelProviderInput / 1MOutput / 1MContextSource
Ling 3.2 Flash (free)inclusionAI$0 (OpenRouter free route)$0 (OpenRouter free route)1M tokensOfficial page
Nov 14, 2026
Perplexity Decider v1.2Perplexity$0.02Free (no output tokens)262K tokens per request (trained on up to about 32K)Official page
Oct 19, 2026
Claude Haiku 5.5Anthropic$0.10 (prompts up to 100K)$0.50 (prompts up to 100K)1M tokens (128K max output)Official page
Oct 15, 2026
Cloudflare ClairicCloudflare$0.10Free (no output tokens)65,536 tokensOfficial page
Oct 20, 2026
GPT-6 LunaOpenAI$0.10$0.501.05M tokens (128K max output)Official page
Oct 15, 2026
MiniMax M3.1 FlashMiniMax$0.15$0.601M tokensOfficial page
Oct 25, 2026
Mercury 3Inception$0.20 ($0.10 at launch)$0.70 ($0.35 at launch)1M tokens (131,072 max output)Official page
Oct 20, 2026
Ripple-65B PreviewRicursive$0.20 (OpenRouter)$0.60 (OpenRouter)1,048,576 tokensOfficial page
Nov 10, 2026
Reflection Beam 1.0ReflectionFrom $0.25 (OpenRouter)From $1.00 (OpenRouter)1M tokens (128K max output)Official page
Oct 17, 2026
DeepSeek V4.1 FlashDeepSeek$0.30 peak / $0.15 off-peak$1.20 peak / $0.60 off-peak1M tokens (384K max output)Official page
Oct 17, 2026
LongCat-2.5Meituan$0.30 (limited-time discount)$1.20 (limited-time discount)1,048,576 tokens (262,144 max output)Official page
Nov 2, 2026
Elorian GlimpseElorian$0.60$3.00262K tokensOfficial page
Oct 17, 2026
Mistral Large 4Mistral$0.70 ($0.07 cached)$2.101M tokensOfficial page
Oct 30, 2026
Gemini 3.8 FlashGoogle$0.75 (intro)$3.75 (intro)1,048,576 tokens (65,536 max output)Official page
Oct 15, 2026
Hy4Tencent$0.834 list ($0.417 at 50% off on TokenHub through Dec 31, 2026)$2.501 list ($1.25 at 50% off on TokenHub through Dec 31, 2026)1,048,576 tokens (64K max output on OpenRouter)Official page
Nov 4, 2026
GPT-6.1 TerraOpenAI$1.00$5.001.05M tokens (922K max input, 128K max output)Official page
Oct 27, 2026
Muse Spark 1.3Meta$1.25$4.251,048,576 tokensOfficial page
Oct 15, 2026
Muse Spark OpenMeta$1.25 (Ollama Cloud)$4.25 (Ollama Cloud)1M tokensOfficial page
Oct 16, 2026
DeepSeek V4 ProDeepSeek$1.32 peak / $0.66 off-peak$3.96 peak / $1.98 off-peak1M tokens (384K max output)Official page
Oct 15, 2026
GLM-5.3Zhipu (Z.ai)$1.40$4.401M tokens (128K max output)Official page
Oct 15, 2026
Gemini 4 HydrogenGoogle$1.50$7.501M tokens (unified input and output)Official page
Nov 13, 2026
Claude Sonnet 5.5Anthropic$2.00$10.001M tokens (128K max output)Official page
Oct 15, 2026
Gemini 4 ArgonGoogle$2.00 (intro, $4.00 from Dec 1)$10.00 (intro, $20.00 from Dec 1)1M tokens (unified input and output)Official page
Oct 15, 2026
Gemini 4 CarbonGoogle$2.00 (intro), then $4.00$10.00 (intro), then $20.00Not published yetOfficial page
Nov 11, 2026
GPT-6.1 SolOpenAI$2.00$10.001.05M tokens (128K max output)Official page
Oct 15, 2026
Grok 4.7xAI$2.00$6.00500K tokensOfficial page
Oct 15, 2026
Kimi K3.1 ProMoonshot$2.00$10.001,048,576 tokensOfficial page
Nov 5, 2026
Qwen3.8 MaxAlibaba Qwen$2.00$6.001M tokensOfficial page
Oct 15, 2026
Kimi K3Moonshot$3.00$15.001,048,576 tokensOfficial page
Oct 15, 2026
Claude Opus 5.5Anthropic$4.00$20.001M tokens (128K max output)Official page
Oct 16, 2026
Claude Fable 5.2Anthropic$8.00$40.001M tokensOfficial page
Nov 10, 2026
Claude Fable 5.1Anthropic$10.00$50.001M tokens (128K max output)Official page
Oct 15, 2026
GPT-6 AstraOpenAI$10.00$50.001.05M tokens (922K max input, 128K max output)Official page
Oct 15, 2026
AlphaGenome 2Google DeepMindNot publishedNot publishedDNA sequences up to 1 million lettersOfficial page
Nov 12, 2026
Cohere Command A+CohereFree on Cohere API (rate limited)Free on Cohere API (rate limited)128K tokens (64K max output)Official page
Oct 15, 2026
GPT-6 CyberOpenAINot publishedNot publishedNot publishedOfficial page
Nov 7, 2026
PAN 2Institute of Foundation ModelsNot publishedNot publishedNot publishedOfficial page
Nov 12, 2026
Seedream 5.5ByteDance SeedNot publishedNot publishedNot publishedOfficial page
Nov 10, 2026

Prices can change. Check the official page before you bill a customer against a row.

14-Day Price Alert

Email me when these prices change

Get notified for the next 14 days whenever an AI lab changes its token pricing or discounts expire.

Free 14-day price change alerts. Zero spam, unsubscribe anytime in one click.

FAQ

How much does an AI model API cost per million tokens?

It depends on the model. In this table, input prices run from $0 (OpenRouter free route) up to GPT-6 Astra at $10.00 in and $50.00 out. Output is usually several times the input price. Local models are not in this table; they are free to run on your own hardware.

Which AI model API is cheapest?

Sorted by the first dollar amount in the vendor's input price, Ling 3.2 Flash (free) is at the top of this table ($0 (OpenRouter free route) in, $0 (OpenRouter free route) out). Some rows are intro or launch-sale prices that expire; read the source link.

What is the context window on current AI APIs?

Most flagship APIs in this table list around 1 million tokens of context. A few are smaller (Cohere Command A+ at 128K, xAI Grok 4.7 at 500K, Cloudflare Clairic at 65,536). The context column is the vendor string, not a number we computed.

Where do these API prices come from?

Each row links to the official page in source_url for that model, the same page we use on the model profile. We do not convert currencies, guess list prices, or blend host markups into a single number unless the vendor page itself states that figure.

Sources

Want tested skills and personas installed in one command? See the AI Loadout packs, from $9.