Pricing

Simple, usage-based pricing. Pay only for what you use.

FastMetal bills per token from a prepaid balance — in Japanese yen, or in US dollars on a US-dollar account — topped up by card, or by bank transfer for yen: each model has its own input and output price per million tokens (below), there is no subscription or minimum, and the free models cost nothing.

Prices are shown in the currency you pick here. You are billed in your account's currency.

Model catalog & pricing

autoSmart router

routes each request to the model that is cheapest for its size among those supporting it (tools/vision/JSON) — billed at the resolved model's price; the listed price is what a typical small request resolves to

Dynamic pricingBilled at the resolved model's rate

Text models

priced per 1M input / output tokens103 models
Claude Fable 5 - Anthropic's flagship Mythos-class model
Input$10.55/ 1M
Output$52.75/ 1M
Claude Fable 5.1 - Anthropic's flagship, stronger agentic coding and long-horizon work than Fable 5
Input$10.55/ 1M
Output$52.75/ 1M
anthropic logoAnthropicanthropic-claude-haiku-4-5🇯🇵Vision
Claude Haiku 4.5 - Fastest with near-frontier intelligence
Input$1.17/ 1M
Output$5.81/ 1M
Claude Haiku 5.5 - Anthropic's fastest, lowest-cost model, succeeding Haiku 4.5
Input$0.1055/ 1M
Output$0.5275/ 1M
Claude Opus 4.6 - Most intelligent, best for agents and coding
Input$5.42/ 1M
Output$26.77/ 1M
Claude Opus 4.8 - Most intelligent, best for agents and coding
Input$5.28/ 1M
Output$26.38/ 1M
Claude Opus 5 - Most intelligent, best for agents and coding
Input$5.28/ 1M
Output$26.38/ 1M
Claude Opus 5.5 - Anthropic's most capable model, priced below Opus 5
Input$4.22/ 1M
Output$21.10/ 1M
anthropic logoAnthropicanthropic-claude-sonnet-4-6🇯🇵Vision
Claude Sonnet 4.6 - Best balance of speed and intelligence
Input$3.49/ 1M
Output$17.41/ 1M
Claude Sonnet 5 - Best balance of speed and intelligence
Input$2.11/ 1M
Output$10.55/ 1M
Claude Sonnet 5.5 - Anthropic's everyday model, succeeding Sonnet 5
Input$2.11/ 1M
Output$10.55/ 1M
Reflection AIbeam-1.0
Reflection Beam 1.0 - Reflection AI's open-weight 501B MoE reasoning model for coding and agentic work, with a 1M-token context
Input$0.422/ 1M
Output$1.27/ 1M
DeepSeek's V4 Flash
Input$0.095/ 1M
Output$0.1899/ 1M
DeepSeek's V4 Flash 0731
Input$0.1477/ 1M
Output$0.2954/ 1M
DeepSeek's V4 Pro
Input$1.38/ 1M
Output$2.75/ 1M
deepseek logoDeepSeekdeepseek-v4.1-flashVision
DeepSeek's V4.1 Flash
Input$0.211/ 1M
Output$0.633/ 1M
DeepSeek's V4.1 Flash 1017 - the October 17, 2026 snapshot, 1M context and image input
Input$0.211/ 1M
Output$0.633/ 1M
DeepSeek V4.1 Flash, uncensored - a fine-tune without the usual refusals, 1M context
Input$0.3482/ 1M
Output$1.40/ 1M
Mistral Devstral 2512
Input$0.44/ 1M
Output$2.20/ 1M
Fireworksember-1Vision
Fireworks Ember-1 - a reasoning model built on Kimi K3
Input$3.30/ 1M
Output$16.50/ 1M
Sakana AIfugu-maxVision
Sakana AI Fugu Max - cost-performance multi-agent orchestration model
Input$2.20/ 1M
Output$6.60/ 1M
Sakana AIfugu-ultra-v2Vision
Sakana AI Fugu Ultra v2 - high-performance multi-agent orchestration model
Input$5.50/ 1M
Output$33.00/ 1M
Google's Gemini 3.1 Pro (preview)
Input$2.11/ 1M
Output$12.66/ 1M
google logoGooglegemini-3.5-flashVision
Google's Gemini 3.5 Flash
Input$1.65/ 1M
Output$9.90/ 1M
google logoGooglegemini-3.7-flashVision
Google's Gemini 3.7 Flash
Input$0.7913/ 1M
Output$3.96/ 1M
google logoGooglegemini-3.8-flashVision
Google's Gemini 3.8 Flash
Input$0.7913/ 1M
Output$3.96/ 1M
google logoGooglegemini-4-argonVision
Google Gemini 4 Argon - Google's flagship reasoning model for software engineering, math and science, and multi-step agents
Input$2.11/ 1M
Output$10.55/ 1M
google logoGooglegemini-flash-lite-freeFreeVision
Google's Gemini Flash Lite - free to try
Input$0.00/ 1M
Output$0.00/ 1M
google logoGooglegemma-4-26b-a4b-itVision
Google Gemma 4 26B-A4B
Input$0.0739/ 1M
Output$0.3587/ 1M
z-ai logoZ.aiglm-4.6vVision
Z.ai GLM 4.6V
Input$0.33/ 1M
Output$0.99/ 1M
Z.ai's GLM 4.7
Input$0.422/ 1M
Output$1.85/ 1M
Z.ai's GLM 4.7 Flash
Input$0.0666/ 1M
Output$0.44/ 1M
z-ai logoZ.aiglm-5
Z.ai's GLM-5
Input$1.06/ 1M
Output$3.38/ 1M
Z.ai's GLM 5.2
Input$1.48/ 1M
Output$4.65/ 1M
Z.ai's GLM 5.3
Input$1.48/ 1M
Output$4.65/ 1M
z-ai logoZ.aiglm-5.3-flashVision
Z.ai GLM 5.3 Flash
Input$0.1583/ 1M
Output$0.5275/ 1M
z-ai logoZ.aiglm-5.3-flashxVision
Z.ai GLM 5.3 FlashX
Input$0.407/ 1M
Output$1.38/ 1M
openai logoOpenAIgpt-4.1-miniVision
OpenAI's GPT-4.1 mini
Input$0.44/ 1M
Output$1.76/ 1M
openai logoOpenAIgpt-4.1-nanoVision
OpenAI's GPT-4.1 nano
Input$0.11/ 1M
Output$0.44/ 1M
openai logoOpenAIgpt-4o-miniVision
OpenAI's GPT-4o mini
Input$0.1583/ 1M
Output$0.633/ 1M
openai logoOpenAIgpt-5-miniVision
OpenAI's GPT-5 mini
Input$0.2638/ 1M
Output$2.11/ 1M
openai logoOpenAIgpt-5.4-nanoVision
OpenAI's GPT-5.4 nano
Input$0.211/ 1M
Output$1.32/ 1M
openai logoOpenAIgpt-5.6-lunaVision
OpenAI's GPT-5.6 Luna
Input$0.211/ 1M
Output$1.27/ 1M
openai logoOpenAIgpt-5.6-solVision
OpenAI's GPT-5.6 Sol
Input$4.40/ 1M
Output$22.00/ 1M
openai logoOpenAIgpt-5.6-terraVision
OpenAI's GPT-5.6 Terra
Input$2.11/ 1M
Output$12.66/ 1M
openai logoOpenAIgpt-6-astraVision
OpenAI's GPT-6 Astra
Input$10.55/ 1M
Output$52.75/ 1M
openai logoOpenAIgpt-6-astra-proVision
OpenAI's GPT-6 Astra Pro
Input$11.00/ 1M
Output$55.00/ 1M
openai logoOpenAIgpt-6-lunaVision
OpenAI GPT-6 Luna
Input$0.1055/ 1M
Output$0.5275/ 1M
OpenAI GPT-6 Luna Pro
Input$0.11/ 1M
Output$0.55/ 1M
openai logoOpenAIgpt-6-solVision
OpenAI GPT-6 Sol
Input$2.11/ 1M
Output$10.55/ 1M
OpenAI GPT-6 Sol Pro
Input$2.20/ 1M
Output$11.00/ 1M
OpenAI GPT-6.1 Sol
Input$2.11/ 1M
Output$10.55/ 1M
OpenAI GPT-6.1 Sol Pro
Input$2.20/ 1M
Output$11.00/ 1M
openai logoOpenAIgpt-6.1-terraVisionChat only
OpenAI GPT-6.1 Terra - the mid-tier GPT-6.1 model, below GPT-6.1 Sol, 1M context
Input$1.10/ 1M
Output$5.50/ 1M
openai logoOpenAIgpt-6.1-terra-proVisionChat only
OpenAI GPT-6.1 Terra Pro - GPT-6.1 Terra in its higher-quality pro mode
Input$1.10/ 1M
Output$5.50/ 1M
OpenAI gpt-oss-20B
Input$0.0317/ 1M
Output$0.1477/ 1M
openai logoOpenAIgpt-oss-120b🇯🇵
gpt-oss-120b
Input$0.102/ 1M
Output$0.5035/ 1M
x-ai logoxAIgrok-4.5Vision
xAI's Grok 4.5
Input$2.20/ 1M
Output$6.60/ 1M
x-ai logoxAIgrok-4.6Vision
xAI's Grok 4.6
Input$2.20/ 1M
Output$6.60/ 1M
x-ai logoxAIgrok-4.7Vision
xAI's Grok 4.7
Input$2.20/ 1M
Output$6.60/ 1M
tencent logoTencenthy3
Tencent Hy3
Input$0.1477/ 1M
Output$0.6119/ 1M
Thinking MachinesinklingVisionChat only
Thinking Machines' Inkling
Input$1.10/ 1M
Output$4.46/ 1M
google logoGooglejapan-gemma-4-31b🇯🇵
Google's Gemma 4 31B - hosted and run in Japan
Input$0.1657/ 1M
Output$0.6436/ 1M
moonshotai logoMoonshotjapan-kimi-k2.6🇯🇵
Moonshot AI's Kimi K2.6 - hosted and run in Japan
Input$0.4079/ 1M
Output$2.02/ 1M
moonshotai logoMoonshotjapan-kimi-k2.7-code🇯🇵
Moonshot AI's Kimi K2.7 Code - hosted and run in Japan
Input$0.3505/ 1M
Output$3.38/ 1M
qwen logoQwenjapan-qwen3.6-35b🇯🇵
Qwen 3.6 35B-A3B - hosted and run in Japan
Input$0.204/ 1M
Output$1.01/ 1M
moonshotai logoMoonshotkimi-k2.6Vision
Moonshot AI's Kimi K2.6
Input$0.7913/ 1M
Output$3.70/ 1M
moonshotai logoMoonshotkimi-k3Vision
Moonshot AI's Kimi K3
Input$3.17/ 1M
Output$15.83/ 1M
moonshotai logoMoonshotkimi-k3.1-proVision
Moonshot AI's Kimi K3.1 Pro - flagship reasoning model, successor to Kimi K3, 1M context and image input
Input$2.11/ 1M
Output$10.55/ 1M
Liquidlfm-2.5-24b-a2bChat only
Liquid AI LFM2.5-24B-A2B - a fast, low-cost MoE reasoning model with 2B active parameters
Input$0.055/ 1M
Output$0.77/ 1M
inclusionAIling-3.0-flash-vlVision
inclusionAI Ling 3.0 Flash VL - low-cost MoE vision-language model
Input$0.0633/ 1M
Output$0.1899/ 1M
Meta Llama 3.1 8B Instruct
Input$0.0211/ 1M
Output$0.0422/ 1M
Meta Llama 3.2 3B Instruct
Input$0.056/ 1M
Output$0.3685/ 1M
Meta Llama 3.3 70B Instruct
Input$0.1055/ 1M
Output$0.3376/ 1M
Meta Llama 4 Maverick
Input$0.22/ 1M
Output$0.88/ 1M
Japanese language model optimized for instruction following
Input$0.102/ 1M
Output$0.5035/ 1M
Inceptionmercury-2.5
Inception Mercury 2.5 - diffusion LLM, the fastest reasoning model
Input$0.044/ 1M
Output$0.165/ 1M
Inceptionmercury-3Chat only
Inception Mercury 3 - Inception's most capable diffusion LLM, very fast, with a 1M-token context
Input$0.11/ 1M
Output$0.385/ 1M
Xiaomimimo-v2.6-flashVision
Xiaomi's MiMo-V2.6-Flash
Input$0.1477/ 1M
Output$0.2954/ 1M
Xiaomimimo-v2.6-proVision
Xiaomi's MiMo-V2.6-Pro
Input$0.459/ 1M
Output$0.9179/ 1M
Minimax's M2.7
Input$0.3165/ 1M
Output$1.27/ 1M
minimax logoMiniMaxminimax-m3Vision
MiniMax's M3
Input$0.2954/ 1M
Output$1.17/ 1M
minimax logoMiniMaxminimax-m3.1-flashVision
MiniMax's M3.1 Flash - fast, low-cost multimodal model with a 1M-token context
Input$0.1583/ 1M
Output$0.633/ 1M
mistralai logoMistralmistral-large-4VisionChat only
Mistral Large 4 - Mistral AI's frontier multimodal model for reasoning, coding and agents
Input$1.50/ 1M
Output$4.60/ 1M
Mistral's voxtral-mini-3b-2507
Input$0.0528/ 1M
Output$0.0528/ 1M
Meta's Muse Glimmer 30B
Input$0.3693/ 1M
Output$1.59/ 1M
meta logoMetamuse-spark-1.2Vision
Meta Muse Spark 1.2
Input$1.38/ 1M
Output$4.68/ 1M
meta logoMetamuse-spark-1.3Vision
Meta Muse Spark 1.3
Input$1.38/ 1M
Output$4.68/ 1M
Meta Muse Spark 1.3 Contributor
Input$0.11/ 1M
Output$0.22/ 1M
meta logoMetamuse-spark-openVisionChat only
Meta Muse Spark Open - Meta's open-weight, coding-optimized MoE model with a 1M-token context
Input$1.38/ 1M
Output$4.68/ 1M
meta logoMetamuse-wonderVisionChat only
Meta Muse Wonder - Meta's frontier reasoning model for complex agentic tasks, with image, video, audio and PDF input
Input$4.40/ 1M
Output$22.00/ 1M
NVIDIA Nemotron 3 Super 120B-A12B
Input$0.0897/ 1M
Output$0.422/ 1M
NVIDIA Nemotron 3.5 Super VL - NVIDIA's open multimodal MoE reasoning model with image and PDF input
Input$0.211/ 1M
Output$1.06/ 1M
UnbiasedparetoVision
Unbiased Pareto 26.10 - blended multimodal model for research, coding and agentic work, 1M context
Input$0.88/ 1M
Output$3.52/ 1M
Qwen3-Next 80B-A3B Instruct
Input$0.095/ 1M
Output$1.17/ 1M
Qwen3-VL 235B-A22B Instruct
Input$0.211/ 1M
Output$0.9284/ 1M
qwen logoQwenqwen3.6-27bVision
Qwen 3.6 27B
Input$0.3376/ 1M
Output$3.38/ 1M
Qwen 3.7 Max
Input$2.64/ 1M
Output$7.92/ 1M
Qwen 3.8 2.4T-A95B
Input$2.20/ 1M
Output$6.60/ 1M
qwen logoQwenqwen3.8-27bVision
Qwen 3.8 27B
Input$0.211/ 1M
Output$2.64/ 1M
qwen logoQwenqwen3.8-maxVision
Qwen 3.8 Max
Input$2.11/ 1M
Output$6.33/ 1M
Upstage Solar Mini 4
Input$0.055/ 1M
Output$0.22/ 1M
Upstagesolar-pro4
Upstage Solar Pro 4
Input$0.165/ 1M
Output$0.66/ 1M

Image models

priced per image5 models
bytedance-seed logoByteDance Seedbytedance-seedream-4.5Image
Generate images
$0.0456/ img
google logoGooglegoogle-nano-banana-2VisionImage
Google Nano Banana 2 - fast, high-quality image generation
$0.078/ img
Venicelustify-v8Chat onlyImage
Lustify v8 - uncensored image generation
$0.0106/ img
Tongyiz-image-turboImage
Generate realistic images
$0.0108/ img
Tongyiz-image-turbo-uncensoredChat onlyImage
Z-Image Turbo, uncensored - realistic images without the nudity filter
$0.0106/ img

Video models

priced per video1 model
bytedance logoByteDanceSeedance 2.0 FastVideo
8s · 1280x720 · with audio
$1.60/ vid

Decision models

priced per decision1 model
TypesafeTypeSafe Jev 1.13Decision
32,000-token context · up to 8,192 bytes per request
$0.00064/ req

"/ 1M" = per 1,000,000 tokens · 🇯🇵 runs inside Japan · "Free" costs nothing within rate limits · "Vision" accepts image input · image, video and decision models are priced per unit. Bars show relative cost on a log scale. Prices before tax.

Pricing questions

All 8 answers, collapsed. Click a question to open it.

01How is the price of a request calculated?
Each model has an input price and an output price per million tokens, listed in the table above in yen or, with the currency switch, in US dollars; an account is always billed in the currency it was opened in. A request costs its prompt tokens at the input rate plus its generated tokens at the output rate; image and video models are priced per image or per video instead, and structured decisions are a flat price per accepted request that does not change with the number of questions or the length of the body. The amount is deducted from your balance as the request completes.
02Is there a subscription or a monthly minimum?
No. You add credit to your balance by card or bank transfer and requests draw it down; payments are processed through Stripe. When the balance reaches zero, paid requests stop until you add more — the balance is also your spending cap.
03What payment methods are there? Can I pay by bank transfer?
For a yen account: credit card or Japanese bank transfer (furikomi), both prepaid, both from ¥500, both with a qualified invoice. A US-dollar account tops up by card, from $5. Choosing bank transfer issues the invoice immediately with a virtual bank account dedicated to you; it is payable for 14 days, and your balance is credited automatically once the transfer lands, so you do not need to notify us. The sender pays the transfer fee. One open transfer invoice at a time. There is no postpaid month-end invoicing.
04Can I try it without paying?
Yes. The free model (gemini-flash-lite-free) can be called at ¥0 with a fresh account, from the browser chat or the API, within a per-key rate limit. Paid models need a positive balance.
05Do the listed prices include consumption tax?
The per-token prices are before tax. Japanese consumption tax is added when you buy credit, and the checkout shows the tax-inclusive total before you pay.
06Do you issue qualified invoices (適格請求書)?
Yes. FastMetal is a registered qualified invoice issuer (registration number T5010001015990) and issues a qualified invoice for every credit purchase. You can pay by credit card or by bank transfer (prepaid). Choosing bank transfer issues a qualified invoice immediately, carrying a bank account dedicated to your account; credit is added automatically once the transfer is confirmed, so there is no need to notify us. Postpaid invoicing — billing in arrears at month end — is not available.
07Does purchased credit expire?
Yes. Prepaid credit carries an expiry date, set out with the other terms in our commercial transaction disclosure. To keep using the service past it, buy additional credit.
08Is unused credit refunded?
Once the service has started, refunds are not given in principle. Credit can be bought from a small amount, so start small and top up once you know how much you use.

Ways to pay

Credit card or Japanese bank transfer (furikomi), both prepaid, both with a qualified invoice.

Credit card

Buy through Stripe Checkout and the balance is available immediately. From ¥500, or $5 on a US-dollar account.

Bank transfer (prepaid)

We issue a qualified invoice straight away, carrying a virtual bank account dedicated to your company. It is payable for 14 days, and the balance is credited automatically once the transfer lands — you do not need to tell us. The sender pays the transfer fee.

Qualified invoice

Registered issuer, number T5010001015990. Issued on either payment method. There is no postpaid month-end invoicing.

How It Works

Purchase Credits

Add funds to your account via Stripe. Your balance is available immediately.

Use the API

Make requests to any supported model. Costs are calculated per token.

Track Usage

Monitor your spending in real-time. Get detailed analytics on your usage.

Using FastMetal at your company

Billing in yen or US dollars, qualified invoices on yen purchases, and a prepaid balance that caps spend — with the questions procurement usually asks answered in one place.

See the business page

Ready to Start?

Create an account and purchase your first credits.

Get Started