Early index, not launched yet.

What AI tokens really cost, per million.

About 60 API providers price tokens differently, and subscriptions hide their limits in 5-hour windows and weekly caps. Tokker puts every model and every plan in dollars per 1M tokens, with the source beside each number.

As of 5 Oct 2026

Open-weight, 18 offers from 16 providers

Cheapest way to buy GLM-5.3, $ per 1M tokens

Input $0.90, output $4.00 per 1M tokens.

DeepInfra, fp4.

GLM-5.3: cheapest offers by provider, US dollars per 1M tokens
Provider	In	Out	Checked
DeepInfrafp4	$0.90	$4.00	5 Oct 2026 source: api.deepinfra.com
Hugging Face Inference ProvidersStandard	$0.90	$4.00	5 Oct 2026 source: router.huggingface.co
Runwaresecondary source	$1.20	$4.00	5 Oct 2026 source: router.requesty.ai
Z.AI (Zhipu)China site, from CNY	$1.19	$4.18	5 Oct 2026 source: docs.bigmodel.cn

Frontier, 7 offers from 7 providers

Cheapest way to buy Claude Opus 5.5, $ per 1M tokens

Input $2.00, output $10.00 per 1M tokens.

Anthropic, batch API, 50% off, results within 24 h. The list price is the same at 6 providers.

Claude Opus 5.5: cheapest offers by provider, US dollars per 1M tokens
Provider	In	Out	Checked
AnthropicBatch API, 50% off, results within 24 h	$2.00	$10.00	5 Oct 2026 source: platform.claude.com
AnthropicStandard	$4.00	$20.00	5 Oct 2026 source: platform.claude.com
Amazon BedrockStandard	$4.00	$20.00	5 Oct 2026 source: pricing.us-east-1.amazonaws.com
DeepInfraStandard	$4.00	$20.00	5 Oct 2026 source: api.deepinfra.com

Open-weight, 18 offers from 17 providers

Cheapest way to buy DeepSeek V4.1 Flash, $ per 1M tokens

Input $0.15, output $0.60 per 1M tokens.

DeepSeek, off-peak hours, half price.

DeepSeek V4.1 Flash: cheapest offers by provider, US dollars per 1M tokens
Provider	In	Out	Checked
DeepSeekOff-peak hours, half price	$0.15	$0.60	5 Oct 2026 source: api-docs.deepseek.com
SiliconFlowInternational site	$0.15	$0.60	5 Oct 2026 source: siliconflow.com
Runwaresecondary source	$0.15	$0.60	5 Oct 2026 source: router.requesty.ai
DeepInfrafp8	$0.20	$0.60	5 Oct 2026 source: api.deepinfra.com

$ per 1M

Claude Opus 5.5, cheapest at Anthropic: input $4.00, output $20.00 per 1M tokensINOUT
GPT-6 Sol, cheapest at OpenAI: input $2.00, output $10.00 per 1M tokensINOUT
Claude Max 20x subscription: about $0.077 per 1M tokens at full useFULL USE
Gemini 3.1 Pro Preview, cheapest at Google (Gemini API): input $2.00, output $12.00 per 1M tokensINOUT
Claude Sonnet 5.5, cheapest at Anthropic: input $2.00, output $10.00 per 1M tokensINOUT
ChatGPT Pro $200 subscription: about $0.044 per 1M tokens at full useFULL USE
Gemini 3.8 Flash, cheapest at Google (Gemini API): input $0.75, output $3.75 per 1M tokensINOUT
Mistral Large 3, cheapest at Mistral AI: input $0.50, output $1.50 per 1M tokensINOUT
Google AI Pro subscription: about $0.021 per 1M tokens at full useFULL USE
MiniMax-M3, cheapest at DeepInfra: input $0.28, output $1.10 per 1M tokensINOUT
Kimi K3, cheapest at Hugging Face Inference Providers: input $2.70, output $13.50 per 1M tokensINOUT
Copilot Pro+ subscription: about $0.59 per 1M tokens at full useFULL USE
GLM-5.3, cheapest at DeepInfra: input $0.90, output $4.00 per 1M tokensINOUT
DeepSeek V4.1 Flash, cheapest at SiliconFlow: input $0.15, output $0.60 per 1M tokensINOUT
GLM Coding Max subscription: about $0.10 per 1M tokens at full useFULL USE
gpt-oss-120b, cheapest at Runware: input $0.032, output $0.14 per 1M tokensINOUT
gpt-oss-20b, cheapest at DeepInfra: input $0.03, output $0.14 per 1M tokensINOUT
Coding Plan Pro subscription: about $0.026 per 1M tokens at full useFULL USE
Qwen3.7-Flash, cheapest at Alibaba Cloud Model Studio: input $0.030, output $0.119 per 1M tokensINOUT
ModelArk Coding Lite subscription: about $0.039 per 1M tokens at full useFULL USE
GLM-4.7-Flash on Z.AI (Zhipu): freePRICE

62 API providers, 480 offers, 103 plans from 32 vendors, 158 sources. Prices as of 5 Oct 2026.

First snapshot. Once checks repeat, moves will show as ▼ cheaper or ▲ pricier.

Cheapest right now

US dollars per 1M tokens, cheapest provider per model. Every number links to the page it came from.

Frontier models

Frontier models: the cheapest provider for each, US dollars per 1M tokens
Model	Cheapest provider	Input	Output	Cache read	Offers	Source
Claude Opus 5.5	Anthropic	$4.00	$20.00	$0.20	7	checked 5 Oct 2026 (source: platform.claude.com)
GPT-6 Sol	OpenAI	$2.00	$10.00	$0.20	5	checked 5 Oct 2026 (source: developers.openai.com)
Gemini 3.1 Pro Preview	Google (Gemini API)	$2.00	$12.00	$0.20	3	checked 5 Oct 2026 (source: ai.google.dev)

Mid models

Mid models: the cheapest provider for each, US dollars per 1M tokens
Model	Cheapest provider	Input	Output	Cache read	Offers	Source
Claude Sonnet 5.5	Anthropic	$2.00	$10.00	$0.20	7	checked 5 Oct 2026 (source: platform.claude.com)
Gemini 3.8 Flash	Google (Gemini API)	$0.75	$3.75	$0.075	5	checked 5 Oct 2026 (source: ai.google.dev)
Mistral Large 3	Mistral AI	$0.50	$1.50	unknown	1	checked 5 Oct 2026 (source: mistral.ai)
MiniMax-M3	DeepInfrafp8	$0.28	$1.10	$0.056	14	checked 5 Oct 2026 (source: api.deepinfra.com)

Open-weight models

Open-weight models: the cheapest provider for each, US dollars per 1M tokens
Model	Cheapest provider	Input	Output	Cache read	Offers	Source
Kimi K3	Hugging Face Inference Providers	$2.70	$13.50	unknown	18	checked 5 Oct 2026 (source: router.huggingface.co)
GLM-5.3	DeepInfrafp4	$0.90	$4.00	$0.20	18	checked 5 Oct 2026 (source: api.deepinfra.com)
DeepSeek V4.1 Flash	SiliconFlowInternational site	$0.15	$0.60	$0.003	18	checked 5 Oct 2026 (source: siliconflow.com)

Cheap models

Cheap models: the cheapest provider for each, US dollars per 1M tokens
Model	Cheapest provider	Input	Output	Cache read	Offers	Source
gpt-oss-120b	Runwaresecondary source	$0.032	$0.14	$0.032	27	checked 5 Oct 2026 (source: router.requesty.ai)
gpt-oss-20b	DeepInfrabfloat16	$0.03	$0.14	unknown	15	checked 5 Oct 2026 (source: api.deepinfra.com)
Qwen3.7-Flash	Alibaba Cloud Model StudioChina site, from CNY	$0.030	$0.119	$0.006	2	checked 5 Oct 2026 (source: help.aliyun.com)

Free models

Free models: the cheapest provider for each, US dollars per 1M tokens
Model	Cheapest provider	Input	Output	Cache read	Offers	Source
GLM-4.7-Flash	Z.AI (Zhipu)International site	Free	Free	Free	2	checked 5 Oct 2026 (source: docs.z.ai)
Command A+	Cohere	Free	Free	unknown	1	checked 5 Oct 2026 (source: cohere.com)

Free tokens have other costs. Z.AI lists GLM-4.7-Flash as free; Cohere's free keys are trial keys, rate-limited and not for production use.

Cheapest means the lowest blended price, three parts input to one part output. OpenRouter is listed but never ranked cheapest: its headline price can pair one endpoint's input price with another's output price. Prices in other currencies are converted at the ECB rate of 2 Oct 2026.

Subscriptions

Ranked by estimated $ per 1M tokens at full use, for agentic coding: 20k input tokens (90% cached) and 1.5k output per call, four 5-hour windows a day.

Subscription plans ranked by estimated US dollars per 1M tokens at full use
Plan	Price / month	$ / 1M at full use	Runs out first	Limits from	Source
Google AI Pro AI ProGoogle	$19.99	$0.021	Daily cap	Vendor page	checked 5 Oct 2026 (source: geminicli.com)
Synthetic subscription Subscription Pack (1)Synthetic	$30	$0.023	Not published	Vendor page	checked 5 Oct 2026 (source: synthetic.new)
Coding Plan ProAlibaba Cloud (Model Studio, intl)	$50	$0.026	Monthly cap	Vendor page	checked 5 Oct 2026 (source: alibabacloud.com)
ModelArk Coding Plan LiteBytePlus (ByteDance intl)	$20	$0.039	Monthly cap	Secondary source	checked 5 Oct 2026 (source: byteplus.com)
ChatGPT / Codex Pro $200OpenAI	$200	$0.044	Not published	Secondary source	checked 5 Oct 2026 (source: learn.chatgpt.com)
Cerebras Code Code MaxCerebras	$200	$0.056	Daily cap	Vendor page	checked 5 Oct 2026 (source: cerebras.ai)
Claude ProAnthropic	$20	$0.065	Not published	Secondary source	checked 5 Oct 2026 (source: claude.com)
Claude Max 5xAnthropic	$100	$0.067	Not published	Secondary source	checked 5 Oct 2026 (source: support.claude.com)
Cerebras Code Code ProCerebras	$50	$0.069	Daily cap	Vendor page	checked 5 Oct 2026 (source: cerebras.ai)
Claude Max 20xAnthropic	$200	$0.077	Weekly cap	Secondary source	checked 5 Oct 2026 (source: support.claude.com)
Google AI Ultra AI Ultra $100 (5x Pro)Google	$100	$0.077	Daily cap	Vendor page	checked 5 Oct 2026 (source: geminicli.com)
ChatGPT / Codex PlusOpenAI	$20	$0.089	Not published	Vendor page	checked 5 Oct 2026 (source: learn.chatgpt.com)
ChatGPT / Codex Pro $100OpenAI	$100	$0.089	Not published	Secondary source	checked 5 Oct 2026 (source: learn.chatgpt.com)
ChatGPT / Codex Pro $500OpenAI	$500	$0.089	Not published	Secondary source	checked 5 Oct 2026 (source: learn.chatgpt.com)
GLM Coding Plan MaxZ.AI (Zhipu)	$168	$0.10	Weekly cap	Secondary source	checked 10 Oct 2026 (source: docs.z.ai)
ChatGPT / Codex BusinessOpenAI	$25	$0.11	Not published	Vendor page	checked 5 Oct 2026 (source: learn.chatgpt.com)
GLM Coding Plan ProZ.AI (Zhipu)	$80	$0.12	Weekly cap	Secondary source	checked 10 Oct 2026 (source: docs.z.ai)
Google AI Ultra AI Ultra $200 (20x Pro)Google	$200	$0.15	Daily cap	Vendor page	checked 5 Oct 2026 (source: geminicli.com)
GLM Coding Plan LiteZ.AI (Zhipu)	$18	$0.16	Weekly cap	Vendor page	checked 10 Oct 2026 (source: docs.z.ai)
T3 Chat ProT3 Chat	$8	$0.16	Monthly credit	Secondary source	checked 5 Oct 2026 (source: t3.chat)
GitHub Copilot MaxGitHub	$100	$0.53	Monthly credit	Vendor page	checked 5 Oct 2026 (source: docs.github.com)
GitHub Copilot Pro+GitHub	$39	$0.59	Monthly credit	Vendor page	checked 5 Oct 2026 (source: docs.github.com)
GitHub Copilot ProGitHub	$10	$0.70	Monthly credit	Vendor page	checked 5 Oct 2026 (source: docs.github.com)
Warp MaxWarp	$200	$0.88	Monthly credit	Vendor page	checked 5 Oct 2026 (source: warp.dev)
JetBrains AI AI UltimateJetBrains	$30	$0.90	Monthly credit	Vendor page	checked 5 Oct 2026 (source: jetbrains.com)
JetBrains AI AI ProJetBrains	$10	$1.05	Monthly credit	Vendor page	checked 5 Oct 2026 (source: jetbrains.com)
GitHub Copilot BusinessGitHub	$19	$1.05	Monthly credit	Vendor page	checked 5 Oct 2026 (source: docs.github.com)
Warp BuildWarp	$20	$1.05	Monthly credit	Vendor page	checked 5 Oct 2026 (source: warp.dev)
Augment StandardAugment Code	$20	$1.05	Monthly credit	Vendor page	checked 5 Oct 2026 (source: augmentcode.com)
GitHub Copilot EnterpriseGitHub	$39	$1.05	Monthly credit	Vendor page	checked 5 Oct 2026 (source: docs.github.com)
Augment BusinessAugment Code	$100	$1.05	Monthly credit	Vendor page	checked 5 Oct 2026 (source: augmentcode.com)
Zed ProZed Industries	$10	$2.31	Monthly credit	Vendor page	checked 5 Oct 2026 (source: zed.dev)
Bolt.new ProStackBlitz	$25	$2.50	Not published	Vendor page	checked 5 Oct 2026 (source: bolt.new)
Warp BusinessWarp	$50	$2.63	Monthly credit	Vendor page	checked 5 Oct 2026 (source: warp.dev)
Bolt.new TeamsStackBlitz	$30	$3.00	Not published	Vendor page	checked 5 Oct 2026 (source: bolt.new)

No absolute limits published, so no estimate: AWS (Kiro), Alibaba Cloud (Model Studio, intl), Alibaba Cloud Bailian (China), Amp (Sourcegraph), Anthropic, BytePlus (ByteDance intl), Chutes, Cognition (Devin / Windsurf), Cursor (Anysphere), Factory, Featherless AI, GitHub, Google, MiniMax, Mistral AI, Moonshot AI, OpenAI, Perplexity, Replit, Tencent Cloud, Volcengine (ByteDance China), Z.AI (Zhipu) and xAI. Those plans are in the index with their verbatim wording.

Same model, international site and China site

Converted to dollars at the ECB rate, input / output per 1M tokens:

GLM-5.3: api.z.ai $1.40 / $4.40, bigmodel.cn $1.19 / $4.18; the China site is cheaper.
Kimi K3: platform.kimi.ai $3.00 / $15.00, platform.moonshot.cn $2.98 / $14.92; about the same price.
MiniMax-M3: platform.minimax.io $0.30 / $1.20, platform.minimaxi.com $0.313 / $1.25; the international site is cheaper.
DeepSeek V4.1 Flash: BytePlus $0.30 / $1.20, Volcengine $0.298 / $1.19; about the same price.

Subscriptions, decoded

5-hour windows, weekly caps and credit pools, turned into real tokens and a price per 1M at full use. At full use, the coding plans below cost $0.021 to $0.10 per 1M tokens. The same agentic workload through the Claude Sonnet 5.5 API costs $1.05: 10 to 51 times more.

Workload

20k in (90% cached) / 1.5k out per call

Custom

Calls per day Input tokens per call Output tokens per call Cache hit %

Custom workloads are coming with the API.

Agentic coding: 5 plans. Cheapest at full use: Google AI Pro, $0.021 per 1M.

Claude Max 20x, effective price at full use

$0.077 per 1M tokens

Claude Max 20x ≈ $0.077 per 1M tokens at full use; the weekly cap runs out before the 5-hour windows do.

Assumes 20k in (90% cached) / 1.5k out per call, four 5-hour windows a day, every cap used. The same workload on the Claude Sonnet 5.5 API costs $1.05 per 1M, 14× more. Based on secondary measurements, not published limits.

$ per 1M at full use, log scale. Pick a plan.

Claude Sonnet 5.5 API$1.05

Plans, up to 6 side by side

Claude Max 20xAnthropic, USUS ChatGPT Pro $200OpenAI, USUS Google AI ProGoogle, US (Gemini CLI)US Copilot Pro+GitHub, USUS GLM Coding MaxZ.AI, internationalCN Coding Plan ProAlibaba Cloud, internationalCN ModelArk Coding LiteBytePlus, internationalCN Cursor ProCursor, USUS

Selected plans at full use for this workload.

Claude Max 20xUS
Anthropic, US
$200 per month
- Per 5-hour windowthe vendor publishes no number
- Per week609M binds first
- Per month2.61B
- $ / 1M at full use$0.077
Above ~190 Mtok a month, this plan beats the API rate.
agentic-coding-v1: 20k in (90% cached) / 1.5k out per call, 4 five-hour windows a day, 30 days.
Vendor’s wording
5-hour window: Max 20x includes 20 times the Pro plan's per-session usage allowance
weekly: Max plans also have a weekly usage limit that applies across all models
checked 5 Oct 2026 (source: support.claude.com)
ChatGPT Pro $200US
OpenAI, US
$200 per month
- Per 5-hour windowno cap at this window
- Per weekthe vendor publishes no number
- Per month4.51B
- $ / 1M at full use$0.044
Above ~190 Mtok a month, this plan beats the API rate.
agentic-coding-v1: 20k in (90% cached) / 1.5k out per call, 4 five-hour windows a day, 30 days.
Vendor’s wording
5-hour window: Pro plans currently have no five-hour limit.
weekly: Weekly limits may also apply.
as written: 20x Plus (10x for new subscribers from 2026-10-30) [secondary: threatfrontier.com / search summaries; official multiple in help.openai.com/en/articles/9793128 which returned 403]
checked 5 Oct 2026 (source: learn.chatgpt.com)
Google AI ProUS
Google, US (Gemini CLI)
$19.99 per month
- Per 5-hour windowthe vendor publishes no number
- Per week226M
- Per month968M binds first
- $ / 1M at full use$0.021
Above ~19.0 Mtok a month, this plan beats the API rate.
agentic-coding-v1: 20k in (90% cached) / 1.5k out per call, 4 five-hour windows a day, 30 days.
Vendor’s wording
daily: Google AI Pro | 1,500 requests [Maximum requests per user per day, Gemini CLI]
5-hour window: 4x higher usage limits; Your limit will now refresh every five hours until you reach your weekly limit. [Gemini app]
checked 5 Oct 2026 (source: geminicli.com)
GLM Coding Maxon GLM-5.3CN
Z.AI, international
$168 per month
- Per 5-hour window74.9M
- Per week374M binds first
- Per month1.60B
- $ / 1M at full use$0.10
Above ~160 Mtok a month, this plan beats the API rate.
agentic-coding-v1: 20k in (90% cached) / 1.5k out per call, 4 five-hour windows a day, 30 days.
Vendor’s wording
5-hour window: Max | 28,000 [5-Hour Credits]
weekly: Max | 140,000 [Weekly Credits]
per request: Model credit usage = (Input tokens x Input multiplier + Cached Input tokens x Cached Input multiplier + Output tokens x Output multiplier) / 10,000 [GLM-5.3: 6.9 / 1.7 / 24]
Fair use: The GLM Coding Plan is strictly limited to use within officially supported tools and products.
checked 10 Oct 2026 (source: docs.z.ai)
Coding Plan ProCN
Alibaba Cloud, international
$50 per month
- Per 5-hour window129M
- Per week968M
- Per month1.94B binds first
- $ / 1M at full use$0.026
Above ~47.6 Mtok a month, this plan beats the API rate.
agentic-coding-v1: 20k in (90% cached) / 1.5k out per call, 4 five-hour windows a day, 30 days.
Vendor’s wording
5-hour window: Up to 6,000 requests per 5 hours
weekly: Up to 45,000 requests per week
monthly: Up to 90,000 requests per month
checked 5 Oct 2026 (source: alibabacloud.com)

How it stays true

Four steps for every number. Step 1 is done for the seed index; the daily re-check and the review are being built.

Sourced

Every price links to the provider's own page. If a provider doesn't publish a number, the index says “unknown”. Nothing is guessed.

GLM-5.3 on api.z.ai: $1.40 in, $4.40 out, from docs.z.ai
Re-checked daily

Each of the 158 sources has a written fetch recipe. Scheduled agents will re-read every provider once a day, and a missed check will mark the row stale.

Planned. Not running yet. Re-checks are paused, so the prices on this page are the 5 Oct 2026 snapshot.
Reviewed change

When a page says something new, the agent will open a change. A person reviews it before the new price goes live.

Planned, as pull requests on the open dataset.
Checked stamp

Each number shows the date it was last checked against its source, everywhere it appears.

$1.40 per 1M input, checked 5 Oct 2026

For developers and agents

Today: the whole dataset as JSON and CSV on GitHub, and llms.txt. Planned: a free JSON API and an MCP server. The responses below are built from today's data.

Prices for one model (planned)

$ curl https://api.tokker.dev/v1/offers?model=claude-opus-5.5

{
  "model": "claude-opus-5.5",
  "unit": "usd_per_1m_tokens",
  "offers": [
    {
      "provider": "anthropic",
      "input": 4.0,
      "output": 20.0,
      "cache_read": 0.2,
      "batch_discount_pct": 50,
      "source_url": "https://platform.claude.com/docs/en/about-claude/pricing.md",
      "last_verified_at": "2026-10-05"
    },
    {
      "provider": "google-vertex",
      "input": 4.0,
      "output": 20.0,
      "cache_read": 0.2,
      "source_url": "https://cloud.google.com/vertex-ai/generative-ai/pricing",
      "last_verified_at": "2026-10-05"
    }
  ]
}

MCP server (planned)

{
  "mcpServers": {
    "tokker": { "url": "https://tokker.dev/mcp" }
  }
}

// tool call
cheapest_provider({ "model": "glm-5.3" })

// result
{
  "provider": "deepinfra",
  "input": 0.9,
  "output": 4.0,
  "source_url": "https://api.deepinfra.com/models/list",
  "last_verified_at": "2026-10-05"
}

Read the data schema

Price-drop alerts

Planned. Pick a model or plan, and get one email when it gets cheaper anywhere.

What AI tokens really cost, per million.

Cheapest right now

Frontier models

Mid models

Open-weight models

Cheap models

Free models

Subscriptions

Same model, international site and China site

Subscriptions, decoded

Claude Max 20xUS

ChatGPT Pro $200US

Google AI ProUS

GLM Coding Maxon GLM-5.3CN

Coding Plan ProCN

How it stays true

Sourced

Re-checked daily

Reviewed change

Checked stamp

For developers and agents

Prices for one model (planned)

MCP server (planned)

Price-drop alerts