AiGanak
Scope a project
LLM Price Index · October 2026Updated weekly

What $100 buys on every LLM API

$100 buys 204,081 support answers on GPT-6 Luna, and 2,040 on GPT-6 Astra. That is a 100× gap for the same task. Here is every current OpenAI, Anthropic and Google model, priced on three everyday tasks.

In short

The AiGanak LLM Price Index shows how many chat replies, support answers and document summaries USD 100 buys on each current OpenAI, Anthropic and Google model at standard list prices, using stated token counts, re-checked weekly, with announced price changes.

See the index →Price your own traffic
Tasks per $100 · standard list prices · checked 26 October 2026

How far $100 goes

More is cheaper. Bars use a log scale so every model stays visible.

Chat reply: 2,000 input tokens, none cached, and a 300-token reply.

  1. 1 GPT-6 LunaOpenAI · Small 285,714$0.00035 each
  2. 2 Claude Haiku 5.5Anthropic · Small 285,714$0.00035 each
  3. 3 Gemini 3.1 Flash-LiteGoogle · Small 105,263$0.00095 each
  4. 4 Gemini 3.5 Flash-LiteGoogle · Small 74,074$0.0014 each
  5. 5 Gemini 3.8 FlashGoogle · Mid 38,095$0.0026 each
  6. 6 Claude Haiku 4.5Anthropic · Small · Previous generation 28,571$0.0035 each
  7. 7 GPT-6.1 SolOpenAI · Mid 14,285$0.0070 each
  8. 8 Claude Sonnet 5.5Anthropic · Mid 14,285$0.0070 each
  9. 9 Gemini 4 ArgonGoogle · Flagship 14,285$0.0070 each
  10. 10 Gemini 3.1 Pro (preview)Google · Flagship · Preview 13,157$0.0076 each
  11. 11 Claude Opus 5.5Anthropic · Flagship 7,142$0.0140 each
  12. 12 GPT-6 AstraOpenAI · Flagship 2,857$0.0350 each
  13. 13 Claude Fable 5.1Anthropic · Flagship 2,857$0.0350 each

Support answer: 6,000 input tokens of which 4,000 (instructions and help text) are cached, and a 500-token answer.

  1. 1 GPT-6 LunaOpenAI · Small 204,081$0.00049 each
  2. 2 Claude Haiku 5.5Anthropic · Small 204,081$0.00049 each
  3. 3 Gemini 3.1 Flash-LiteGoogle · Small 74,074$0.0014 each
  4. 4 Gemini 3.5 Flash-LiteGoogle · Small 50,761$0.0020 each
  5. 5 Gemini 3.8 FlashGoogle · Mid 27,210$0.0037 each
  6. 6 Claude Haiku 4.5Anthropic · Small · Previous generation 20,408$0.0049 each
  7. 7 GPT-6.1 SolOpenAI · Mid 10,638$0.0094 each
  8. 8 Claude Sonnet 5.5Anthropic · Mid 10,638$0.0094 each
  9. 9 Gemini 4 ArgonGoogle · Flagship 10,638$0.0094 each
  10. 10 Gemini 3.1 Pro (preview)Google · Flagship · Preview 9,259$0.0108 each
  11. 11 Claude Opus 5.5Anthropic · Flagship 5,319$0.0188 each
  12. 12 Claude Fable 5.1Anthropic · Flagship 2,173$0.0460 each
  13. 13 GPT-6 AstraOpenAI · Flagship 2,040$0.0490 each

Document summary: A 20,000-token document in, no caching, and an 800-token summary.

  1. 1 GPT-6 LunaOpenAI · Small 41,666$0.0024 each
  2. 2 Claude Haiku 5.5Anthropic · Small 41,666$0.0024 each
  3. 3 Gemini 3.1 Flash-LiteGoogle · Small 16,129$0.0062 each
  4. 4 Gemini 3.5 Flash-LiteGoogle · Small 12,500$0.0080 each
  5. 5 Gemini 3.8 FlashGoogle · Mid 5,555$0.0180 each
  6. 6 Claude Haiku 4.5Anthropic · Small · Previous generation 4,166$0.0240 each
  7. 7 GPT-6.1 SolOpenAI · Mid 2,083$0.0480 each
  8. 8 Claude Sonnet 5.5Anthropic · Mid 2,083$0.0480 each
  9. 9 Gemini 4 ArgonGoogle · Flagship 2,083$0.0480 each
  10. 10 Gemini 3.1 Pro (preview)Google · Flagship · Preview 2,016$0.0496 each
  11. 11 Claude Opus 5.5Anthropic · Flagship 1,041$0.0960 each
  12. 12 GPT-6 AstraOpenAI · Flagship 416$0.2400 each
  13. 13 Claude Fable 5.1Anthropic · Flagship 416$0.2400 each
Announced changes

Where $100 will buy less

Google has published dated price increases. Support answers per $100, before and after.

What the Gemini increase means for your bill →

Gemini 4 Argon

10,638 today → 5,319 from 1 December 2026

Gemini 3.8 Flash

27,210 today → 13,605 from 1 January 2027

Method

How the index is calculated

Each figure is USD 100 divided by the cost of one task at the provider's standard pay-as-you-go list price, rounded down. Uncached input is priced at the input rate, cached input at the cached rate, and output, including any thinking tokens, at the output rate. Long-prompt surcharges apply where a task crosses a model's threshold. Batch discounts, cache-write charges, storage fees, taxes and negotiated discounts are not included.

Every model is given the same token counts, which is the main simplification: in practice, models use different numbers of tokens for the same task, and thinking tokens can add a lot. Prices are copied from each provider's own pricing page and re-checked every week: OpenAI, Anthropic, Google.

Price is not quality. A cheaper model only saves money if it does your task well enough; test on your own examples before switching.

Cite this index

Free to quote and republish with a link. Suggested citation:

AiGanak LLM Price Index (October 2026), https://aiganak.com/llm-price-index/
FAQ

LLM Price Index questions

Can't find it? Press ⌘K and ask.

What is the AiGanak LLM Price Index?

A weekly comparison of how many chat replies, support answers and document summaries USD 100 buys on each current OpenAI, Anthropic and Google model at standard list prices. Every model gets the same stated token counts so the arithmetic can be checked.

Which LLM API gives the most for $100?

At list prices checked by AiGanak, GPT-6 Luna and Claude Haiku 5.5 give the most tasks per USD 100 on all three tasks in the index. The cheapest model is only the right choice if it handles your task well.

Why can real costs differ from the index?

The index gives every model the same token counts. In practice, models tokenize text differently and some use many thinking tokens, which are billed as output. Batch discounts, cache-write charges and negotiated prices are also left out.

Can I cite or republish the index?

Yes. It is free to quote and republish with a link to aiganak.com/llm-price-index/.

Your traffic is not our three tasks.

Enter your own request volume and token counts and see every model's monthly cost, or drop in your usage export to see where your money goes.

Open the pricing calculator →Free cost check