Skip to content

Categories

Fordeal

AI API prices per million tokens

Input and output token prices of OpenAI, Anthropic, Google, xAI, Mistral and DeepSeek models, converted at the ČNB rate, with a calculator: how much X messages a month cost, and when a subscription is cheaper.

Prices from official price lists, converted at the Czech National Bank rate of 02/10/2026. We check the price lists every week; a person confirms every change.

How much do I pay for the API?

Enter how many messages you send a month. A typical message is 1 500 tokens in (your question with context) and 600 tokens out (about 1 100 and 450 words). The table converts at the ČNB rate and compares with each provider's cheapest paid subscription.

ModelA monthA messageSubscription or API?
GPT-6 Luna APIOpenAI148 HUF0,15 HUFAPI is cheaperChatGPT Go at 2 631 HUF pays off from 17 774 messages a month
Mistral Small 4 APIMistral192 HUF0,19 HUFNo chat subscription from this provider
DeepSeek V4.1 Flash APIDeepSeek385 HUF0,38 HUFNo chat subscription from this provider
Mistral Large 3 APIMistral543 HUF0,54 HUFNo chat subscription from this provider
Gemini 3.5 Flash-Lite APIGoogle641 HUF0,64 HUFNo chat subscription from this provider
Grok Build 0.1 APIxAI888 HUF0,89 HUFAPI is cheaperGrok SuperGrok at 9 868 HUF pays off from 11 111 messages a month
Gemini 3.8 Flash APIGoogle1 110 HUF1,11 HUFNo chat subscription from this provider
Grok 4.3 APIxAI1 110 HUF1,11 HUFAPI is cheaperGrok SuperGrok at 9 868 HUF pays off from 8 888 messages a month
DeepSeek V4 Pro APIDeepSeek1 433 HUF1,43 HUFNo chat subscription from this provider
Claude Haiku 4.5 APIAnthropic1 480 HUF1,48 HUFAPI is cheaperClaude Pro at 6 646 HUF pays off from 4 489 messages a month
Grok 4.7 APIxAI2 171 HUF2,17 HUFAPI is cheaperGrok SuperGrok at 9 868 HUF pays off from 4 545 messages a month
Mistral Medium 3.5 APIMistral2 220 HUF2,22 HUFNo chat subscription from this provider
Claude Sonnet 5.5 APIAnthropic2 960 HUF2,96 HUFAPI is cheaperClaude Pro at 6 646 HUF pays off from 2 244 messages a month
GPT-6.1 Sol APIOpenAI2 960 HUF2,96 HUFSubscription is cheaperChatGPT Go at 2 631 HUF pays off from 888 messages a month
Gemini 3.1 Pro (Preview) APIGoogle3 355 HUF3,36 HUFNo chat subscription from this provider
Claude Opus 5.5 APIAnthropic5 921 HUF5,92 HUFAPI is cheaperClaude Pro at 6 646 HUF pays off from 1 122 messages a month
Claude Fable 5.1 APIAnthropic14 802 HUF14,8 HUFSubscription is cheaperClaude Pro at 6 646 HUF pays off from 448 messages a month
GPT-6 Astra APIOpenAI14 802 HUF14,8 HUFSubscription is cheaperChatGPT Go at 2 631 HUF pays off from 177 messages a month

  • 1. OpenAI

    GPT-6 Luna API

    Input, 1M tokens
    32,89 HUF
    0,10 US$
    Output, 1M tokens
    164 HUF
    0,50 US$
    Context: 1 050 000
    • Cached input $0.01 / 1M tokens
    • Cheapest, most efficient tier

    Same >272K long-context caveat. Previous gpt-5.4-mini ($0.75/$4.50) and gpt-5.4-nano ($0.20/$1.25) are still listed.

    Source: developers.openai.com, checked 05/10/2026

  • 2. Mistral

    Mistral Small 4 API

    Input, 1M tokens
    49,34 HUF
    0,15 US$
    Output, 1M tokens
    197 HUF
    0,60 US$
    • Cached input −90% (generic discount)

    Context window not shown on the pricing page.

    Source: mistral.ai, checked 05/10/2026

  • 3. DeepSeek

    DeepSeek V4.1 Flash API

    Input, 1M tokens
    98,68 HUF
    0,30 US$
    Output, 1M tokens
    395 HUF
    1,20 US$
    Context: 1 000 000
    • Cached input $0.006 / 1M tokens (peak)
    • Off-peak half price: input $0.15, output $0.60
    • Max output 384K tokens

    Peak price. Off-peak: input $0.15, cached $0.003, output $0.60.

    Source: api-docs.deepseek.com, checked 05/10/2026

  • 4. Mistral

    Mistral Large 3 API

    Input, 1M tokens
    164 HUF
    0,50 US$
    Output, 1M tokens
    493 HUF
    1,50 US$
    • Open-weight model
    • Cached input −90% (no per-model price shown)
    • Batch at half price

    Context window not shown on the pricing page.

    Source: mistral.ai, checked 05/10/2026

  • 5. Google

    Gemini 3.5 Flash-Lite API

    Input, 1M tokens
    98,68 HUF
    0,30 US$
    Output, 1M tokens
    822 HUF
    2,50 US$
    • Cached input $0.03 / 1M tokens
    • Stable release

    Context window not read on an official page.

    Source: ai.google.dev, checked 05/10/2026

  • 6. xAI

    Grok Build 0.1 API

    Input, 1M tokens
    329 HUF
    1 US$
    Output, 1M tokens
    658 HUF
    2 US$
    Context: 256 000
    • Cached input $0.20 / 1M tokens
    • Prompts from 200k: input $2 / output $4

    Cheapest xAI model listed; positioning (coding?) not confirmed. From 200k tokens cached input is $0.40.

    Source: docs.x.ai, checked 05/10/2026

  • 7. Google

    Gemini 3.8 Flash API

    Input, 1M tokens
    247 HUF
    0,75 US$
    Output, 1M tokens
    1 233 HUF
    3,75 US$
    Context: 1 048 576
    • Cached input $0.075 / 1M tokens
    • Max output 65,536 tokens

    Promotional price until 31 Dec 2026; from 1 Jan 2027 input $1.50, output $7.50, caching $0.15.

    Source: ai.google.dev, checked 05/10/2026

  • 8. xAI

    Grok 4.3 API

    Input, 1M tokens
    411 HUF
    1,25 US$
    Output, 1M tokens
    822 HUF
    2,50 US$
    Context: 1 000 000
    • Cached input $0.20 / 1M tokens
    • Prompts from 200k: input $2.50 / output $5

    From 200k tokens cached input is $0.40.

    Source: docs.x.ai, checked 05/10/2026

  • 9. DeepSeek

    DeepSeek V4 Pro API

    Input, 1M tokens
    434 HUF
    1,32 US$
    Output, 1M tokens
    1 303 HUF
    3,96 US$
    Context: 1 000 000
    • Cached input $0.044 / 1M tokens (peak)
    • Off-peak half price: input $0.66, output $1.98
    • Max output 384K tokens

    Peak price (01–04 and 06–10 UTC Mon–Fri). Off-peak (other hours, weekends, Chinese holidays) is half: input $0.66, cached $0.022, output $1.98.

    Source: api-docs.deepseek.com, checked 05/10/2026

  • 10. Anthropic

    Claude Haiku 4.5 API

    Input, 1M tokens
    329 HUF
    1 US$
    Output, 1M tokens
    1 645 HUF
    5 US$
    Context: 200 000
    • Cached input $0.10 / 1M tokens
    • 200k token context

    API id claude-haiku-4-5-20251001. Retirement not sooner than 15 Oct 2026 (watch for a successor).

    Source: platform.claude.com, checked 05/10/2026

  • 11. xAI

    Grok 4.7 API

    Input, 1M tokens
    658 HUF
    2 US$
    Output, 1M tokens
    1 974 HUF
    6 US$
    Context: 500 000
    • Cached input $0.50 / 1M tokens
    • Prompts from 200k: input $4 / output $12

    From 200k tokens the higher rate (cached $1.00) applies to all tokens of the request.

    Source: docs.x.ai, checked 05/10/2026

  • 12. Mistral

    Mistral Medium 3.5 API

    Input, 1M tokens
    493 HUF
    1,50 US$
    Output, 1M tokens
    2 467 HUF
    7,50 US$
    • Frontier-class, agentic and coding
    • Cached input −90% (generic discount)

    More expensive than Mistral Large 3. Context window not shown on the pricing page.

    Source: mistral.ai, checked 05/10/2026

  • 13. Anthropic

    Claude Sonnet 5.5 API

    Input, 1M tokens
    658 HUF
    2 US$
    Output, 1M tokens
    3 289 HUF
    10 US$
    Context: 1 000 000
    • Cached input $0.20 / 1M tokens
    • No surcharge for long context
    • Batch −50%

    Cache write 5 min $2.50, 1 h $4.

    Source: platform.claude.com, checked 05/10/2026

  • 14. OpenAI

    GPT-6.1 Sol API

    Input, 1M tokens
    658 HUF
    2 US$
    Output, 1M tokens
    3 289 HUF
    10 US$
    Context: 1 050 000
    • Cached input $0.10 / 1M tokens
    • Near-Astra performance at lower cost

    Same long-context caveat: input over 272K may be billed at double rates; not confirmed for this model.

    Source: developers.openai.com, checked 05/10/2026

  • 15. Google

    Gemini 3.1 Pro (Preview) API

    Input, 1M tokens
    658 HUF
    2 US$
    Output, 1M tokens
    3 947 HUF
    12 US$
    • Cached input $0.20 / 1M tokens
    • Prompts over 200k: input $4 / output $18

    Preview status. Caching over 200k is $0.40. Context window not read on an official page.

    Source: ai.google.dev, checked 05/10/2026

  • 16. Anthropic

    Claude Opus 5.5 API

    Input, 1M tokens
    1 316 HUF
    4 US$
    Output, 1M tokens
    6 579 HUF
    20 US$
    Context: 1 000 000
    • Cached input $0.20 / 1M tokens
    • No surcharge for long context
    • Fast mode $8 / $40; Batch −50%

    Recommended default. Cache write 5 min $5, 1 h $8.

    Source: platform.claude.com, checked 05/10/2026

  • 17. Anthropic

    Claude Fable 5.1 API

    Input, 1M tokens
    3 289 HUF
    10 US$
    Output, 1M tokens
    16 447 HUF
    50 US$
    Context: 1 000 000
    • Cached input $0.25 / 1M tokens
    • Full 1M context at standard price
    • Batch −50%; cache write 5 min $12.50, 1 h $20

    Top tier above Opus. US-only inference (inference_geo us) costs 1.1x.

    Source: platform.claude.com, checked 05/10/2026

  • 18. OpenAI

    GPT-6 Astra API

    Input, 1M tokens
    3 289 HUF
    10 US$
    Output, 1M tokens
    16 447 HUF
    50 US$
    Context: 1 050 000
    • Cached input $1.00 / 1M tokens
    • Flagship, most capable model

    Pricing page: input over 272K tokens is billed at double input/cache rates for models with long-context pricing; not confirmed for GPT-6 models.

    Source: developers.openai.com, checked 05/10/2026

Some of these services may pay us a commission when you subscribe through our link. It never changes the order: cheapest first, nothing else.