AI Plan & API Cost Calculator
October 2026 — compare subscriptions, API pricing and hybrid setups based on how you actually use AI.
Base pricing last verified 13 October 2026; source-specific dates are listed below. Provider prices, quotas and model availability can change.
October 2026
What's changed in October 2026?
- Claude Haiku 5.5 costs $0.10/$0.50 per M tokens (prompts up to 100K; $0.50/$2.50 above), a tenth of Haiku 4.5. That makes it a strong low-cost choice for classification, summaries and routing. Available now via the Claude API. Opus 5.5 is also cheaper ($4/$20), and Sonnet 5.5 halves cache-read to $0.10/M.
- GPT-6.1 Sol and GPT-6 Luna replace the GPT-5.6 line in the calculator: Sol is $2/$10 (GPT-5.6 Sol was $4/$20 on promotion), Luna is $0.10/$0.50. Available now via the OpenAI API; Plus and Pro use Sol, Free and Go use Luna.
- Meta Muse Wonder (12 October) is a directly priced frontier API at $4/$20 per M, 1M context. Muse Wonder Full is subscription-only, inside Muse Code Power ($50) and Maximum ($200). Meta does not state GA or regional availability, so confirm access first.
- Claude Max now includes monthly API credits ($100 on Max 5x, $200 on Max 20x, from 7 October). They fund API, Agent SDK and
claude -puse with an API key, not interactive Claude Code, and do not roll over. Available now to claim after 7 days on the plan. - Gemini 4 Argon is not in the calculator. It was announced on 30 September but is absent from Google's pricing page and reportedly limited to an early-access programme. Gemini 3.8 Flash ($0.75/$3.75, doubling on 1 January 2027) is used instead. Argon: not purchasable today.
Specialist option, not in the comparison: TypeSafe's Jev 1.13 is a decision-only model. It returns a typed choice, score or yes/no probability, not text, so it only suits classification and routing (lead triage, ticket routing, document categories). Listed at $0.042 per M input tokens and no output charge, with a 64K context. Haiku 5.5 and GPT-6 Luna cost $0.10/M input plus $0.50/M output, so Jev is cheaper on paper for short, structured decisions, but accuracy is task-specific and TypeSafe's own benchmarks are vendor-reported. Test it on your data before switching.
Configurator
Tell us how you actually use AI
Pick a preset to start fast, then adjust anything. Recommendations update instantly. Your setup is saved in the URL so you can share it.
Recommendation
What should you actually use?
Deterministic rules applied to your inputs above, not a ranked affiliate list. See assumptions on each card. Recommendations are drawn from Anthropic, OpenAI and Google — the three platforms with directly comparable coding-capable subscriptions. xAI, Meta, DeepSeek and OpenRouter are shown below for cost and API comparison.
Compare against your current setup (optional)
Provider comparison
Consumer subscriptions vs API pricing
A subscription is a usage allowance, not a token budget — providers don't publish a direct token equivalent for Pro/Plus/Ultra plans. Use the subscription for sustained interactive use, and the API table below for anything you're metering, automating or running in parallel.
Major AI platforms: OpenAI · Anthropic · Google · xAI · Meta. Alternative / multi-model options: DeepSeek · OpenRouter (a multi-model router, not a model vendor — it doesn't build the models it serves).
Consumer subscriptions
Subscription quotas (messages/usage caps) are set by each provider and are not published as a fixed token allowance. Treat the API table as the only ground truth for cost-per-token; treat the subscription table as an interactive-use budget cap.
API pricing
API cost calculator
Raw API cost calculator
monthly cost = (input tokens / 1,000,000 × input price) + (output tokens / 1,000,000 × output price)
Editorial
Subscription vs PAYG: the rule of thumb
- Subscriptions are attractive for sustained, interactive, single-user work — a coding assistant you talk to all day benefits from a flat fee.
- API/PAYG pricing is excellent for bounded, parallel or automated work — background jobs, batch extraction, multiple concurrent agents — where cost tracks exactly with tokens used.
- Hybrid setups — one subscription for the primary interactive assistant, API for everything else — often produce the best cost/capability balance.
- Paying for multiple premium subscriptions (e.g. Claude Max and ChatGPT Pro at once) usually buys duplicated capacity rather than duplicated capability.
- Model routing — sending simple/background work to a small model and reserving frontier models for high-stakes tasks — can materially lower total AI spend.
We don't quote a universal break-even point: it depends on your token volume and concurrency, which is exactly what the calculator above derives from your own inputs.
// Free AI Stack Review
Want help choosing the right AI setup?
nCrypt offers a free 20-minute review of your AI subscriptions, API usage, workflows and automation opportunities — pick a time that works.
Methodology
Pricing sources — base data checked 13 October 2026
Prices shown in USD and exclude taxes. Providers can change pricing, quotas and model availability without notice.