What will this AI work cost?
API-equivalent estimates for every listed model at published rates, calculated in your browser.
Selected: One hour of coding help
Pair-programming passes inside one clock hour
Size
Steady help across a few files. Authored illustration, not a measured average.
USD · API estimate · Rates checked 17 October 2026 ·
Choose an activity
30 authored examples
Model passes during one hour of interactive coding help. The hour is a label for the pass count; it does not predict code shipped or model runtime.
- Model passes
- 20
- Cached prefix per pass
- 20,000 tokens
- Fresh input per pass
- 2,000 tokens
- Output per pass (incl. reasoning)
- 3,000 tokens
- Cache builds
- 1
Not included: Test runner and CI compute; Tools other than web search; Human time.
One instance priced. Context held constant; cache reuse adapts to each model and is explained in its details.
Lowest estimated cost
$0.040GPT-6 LunaClaude Haiku 5.5(tie)
One hour of coding help · Medium · 1 activity
USD · API estimate · Rates checked 17 October 2026 ·
Current models with a complete estimate. Not a quality ranking.
7 current models priced · 4 legacy listed separately
Cache reuse saves $0.034 on GPT-6 Luna under these assumptions.
Model comparison
One hour of coding help · Medium · 1 activity
Current models
| Details | ||
|---|---|---|
| GPT-6 LunaOpenAI | $0.040 | |
| Claude Haiku 5.5Anthropic | $0.040 | |
| Gemini 4 ArgonGoogle | $0.758 | |
| GPT-6.1 SolOpenAI | $0.768 | |
| Claude Sonnet 5.5Anthropic | $0.768 | |
| Claude Opus 5.5Anthropic | $1.54 | |
| GPT-6 AstraOpenAI | $4.03 |
Choose one authored activity and size, or enter one API request in Tokens mode.
Use the dated direct API price snapshot for each listed model, including supported cache and search fees.
Apply the model’s rate rules and limits, then mark estimates that are partial or unsupported.
One hour of coding help · Medium (Steady help across a few files) · one of 2 tied lowest complete current-model estimates: GPT-6 Luna
Rates checked 17 October 2026. Direct API, Standard, global USD prices; subscription pricing is not modeled.
Calculation rules
- An API-equivalent estimate at published list rates, not a bill or a subscription price.
- Standard global direct API prices in USD; batch, regional, priority and enterprise pricing are not modeled.
- Activities price one instance of an authored illustration with constant context. Sizes are examples, not measured averages, benchmarks or a promise that the work succeeds.
- Tokens mode prices exactly one API request. Input includes cached input and cache writes, so ordinary input = input − cached input − cache writes. Output includes reasoning and is counted once.
- A cache write replaces the input price for that bucket; it is not added on top. Cache never transfers between models, so each activity row assumes its own cache builds.
- Price tiers apply per request to all input buckets: Claude Haiku 5.5 above 100K input tokens; every listed OpenAI model above 272K (2× input and cache, 1.5× output for the full request).
- Requests beyond a model's published context or output limit, or with cached tokens below its cache minimum, are marked not supported with the reason.
- Searches are billable search calls, not questions. Retrieved content is already counted in input tokens.
- When a model's search fee is not in this snapshot, its row shows the known subtotal with “Search fees extra”. It is never shown as $0 and is left out of the lowest-cost headline.
- Inputs and results stay in this browser: no API keys, uploads or persistent calculator-data storage. The site records this pageview, without calculation inputs or results.
Not modeled
- Subscriptions, plan allowances, credits, taxes and negotiated contracts
- Batch, Flex, Fast, Ultrafast, priority, regional and US-only processing modifiers
- Tools other than the listed web-search fees: code execution, containers, file search, storage, fetch, image and audio
- Embeddings, vector storage, hosting, test-runner compute and human labor
Source citations
- GPT-6.1 Sol, GPT-6 Astra, GPT-6 Luna, GPT-6 Sol, GPT-5.6 Sol: OpenAI API pricing
- GPT-6.1 Sol, GPT-6 Astra, GPT-6 Luna, GPT-6 Sol, GPT-5.6 Sol: OpenAI prompt caching
- GPT-6.1 Sol, GPT-6 Astra, GPT-6 Luna, GPT-6 Sol, GPT-5.6 Sol: OpenAI web search tool
- Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5, Claude Opus 5, Claude Sonnet 5: Claude API pricing
- Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5, Claude Opus 5, Claude Sonnet 5: Claude prompt caching
- Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5: Claude web search tool
- Gemini 4 Argon: Gemini API pricing
- GPT-6.1 Sol: GPT-6.1 Sol model page
- GPT-6 Astra: GPT-6 Astra model page
- GPT-6 Luna: GPT-6 Luna model page
- GPT-6 Sol: GPT-6 Sol model page
- GPT-5.6 Sol: GPT-5.6 Sol model page
- Claude Opus 5.5: Claude Opus 5.5 model page
- Claude Sonnet 5.5: Claude Sonnet 5.5 model page
- Claude Haiku 5.5: Claude Haiku 5.5 model page
- Claude Opus 5: Claude Opus 5 model page
- Claude Sonnet 5: Claude Sonnet 5 model page
- Gemini 4 Argon: Gemini 4 Argon guide
Search fees are included only where this snapshot records the exact provider API or tool path. Retrieved search content is counted in input tokens; a missing fee remains “Search fees extra.”
JSON includes every applied rate, rule and request group. Files are created in your browser; nothing is uploaded or stored.