Methodology
Much Inference answers one question: for each dollar you spend, how much model output can you get? Every row is a plan (an API or a subscription) paired with a model it gives you access to.
Blended tokens
Input and output tokens are priced differently, so prices are compared on a blended token: 3 input tokens for every 1 output token, a common convention for chat and coding workloads.
Pay-as-you-go APIs
API rows use the price the plan charges per million tokens:
value vs API = model's list blended price ÷ this plan's blended price
A first-party API at list price is 1.0× by definition. Discounted routes, such as batch APIs, score above 1.0×.
Subscriptions
Vendors describe subscription limits in messages, sessions or "x times more usage", not tokens. Each subscription row therefore stores one number: the API-list-price value of a fully used month, meaning what the same usage would cost on the vendor's own API. Tokens follow from the model's list price:
tokens per $1 = tokens per month ÷ monthly price
value vs API = monthly value ÷ monthly price
Where a vendor publishes included usage in dollars (for example, credit-based plans), the row is marked verified. Otherwise it is an est. estimate, and its basis is shown below. Real-world value depends on your workload. Prompt caching, long contexts and reasoning effort all shift the numbers.
Model tiers
Frontier models are each vendor's most capable generally available models, mid are their balanced workhorses, and small are their fastest, cheapest models. Compare within a tier: a small model that is cheaper per token is not doing the same work as a frontier model.
Sources
Every row, its inputs, and where they came from. Prices are monthly billing in USD.
| Row | Inputs | Basis | Checked |
|---|---|---|---|
| OpenAI API · GPT-6 Luna | $0.10 / $0.50 per MTok | List price. Source | 2026-10-06 |
| Google AI Ultra 20x · Gemini 3.1 Proest. | $199.99/mo → $4,000 value | Advertised as 20x Pro usage limits; scaled from the Pro estimate. USD price not confirmed on the official page. Source | 2026-10-06 |
| Claude Max 20x · Claude Opus 5.5est. | $200/mo → $6,000 value | Advertised as 20x Pro usage per 5-hour session; scaled from the Pro estimate. Weekly caps may bind sooner. Source | 2026-10-06 |
| Google AI Pro · Gemini 3.1 Proest. | $19.99/mo → $200 value | Compute-based limits that refresh every 5 hours (since May 2026); no token quota is published. Assumes about $200/mo of Gemini 3.1 Pro at API list price. Source | 2026-10-06 |
| Google AI Ultra · Gemini 3.1 Proest. | $99.99/mo → $1,000 value | Advertised as 5x Pro usage limits; scaled from the Pro estimate. Source | 2026-10-06 |
| ChatGPT Pro ($200) · GPT-5.5est. | $200/mo → $5,000 value | Advertised as 20x Plus usage; scaled from the Plus estimate. Source | 2026-10-06 |
| Claude Max 5x · Claude Opus 5.5est. | $100/mo → $1,500 value | Advertised as 5x Pro usage per 5-hour session; scaled from the Pro estimate. Source | 2026-10-06 |
| Claude Pro · Claude Opus 5.5est. | $20/mo → $300 value | No token quota is published; limits reset every 5 hours with a weekly cap. Assumes a user who hits every limit gets about $300/mo of Opus 5.5 at API list price. Source | 2026-10-06 |
| Gemini API · Gemini 3.1 Flash-Lite | $0.25 / $1.50 per MTok | List price. Source | 2026-10-06 |
| ChatGPT Pro ($100) · GPT-5.5est. | $100/mo → $1,250 value | Launched April 2026 with 5x Plus Codex usage; scaled from the Plus estimate. Source | 2026-10-06 |
| ChatGPT Plus · GPT-5.5est. | $20/mo → $250 value | OpenAI publishes message and Codex caps, not tokens. Assumes about $250/mo of GPT-5.5 at API list price at full utilization. Source | 2026-10-06 |
| Claude Batch API · Claude Haiku 4.5 | $0.50 / $2.50 per MTok | Batch API: 50% off list price for asynchronous jobs. Source | 2026-10-06 |
| Gemini API · Gemini 3.8 Flash | $0.75 / $3.75 per MTok | Promotional price through 2026-12-31; rises to $1.50 / $7.50 on 2027-01-01. Source | 2026-10-06 |
| Claude Batch API · Claude Sonnet 5.5 | $1 / $5 per MTok | Batch API: 50% off list price for asynchronous jobs. Source | 2026-10-06 |
| Claude API · Claude Haiku 4.5 | $1 / $5 per MTok | List price. Source | 2026-10-06 |
| GitHub Copilot Pro+ · Claude Sonnet 5.5 | $39/mo → $70 value | Includes 7,000 AI credits per month (1 credit = $0.01). Assumes credits are spent at the model's API list price. Source | 2026-10-06 |
| GitHub Copilot Pro · Claude Sonnet 5.5 | $10/mo → $15 value | Includes 1,500 AI credits per month (1 credit = $0.01). Assumes credits are spent at the model's API list price. Source | 2026-10-06 |
| Claude Batch API · Claude Opus 5.5 | $2 / $10 per MTok | Batch API: 50% off list price for asynchronous jobs. Source | 2026-10-06 |
| Claude API · Claude Sonnet 5.5 | $2 / $10 per MTok | List price. Source | 2026-10-06 |
| OpenAI API · GPT-6.1 Sol | $2 / $10 per MTok | List price. Source | 2026-10-06 |
| Gemini API · Gemini 3.1 Pro | $2 / $12 per MTok | Preview model; prompts up to 200k tokens ($4/$18 above). Source | 2026-10-06 |
| Claude API · Claude Opus 5.5 | $4 / $20 per MTok | List price. Source | 2026-10-06 |
| OpenAI API · GPT-5.5 | $5 / $30 per MTok | List price. Source | 2026-10-06 |
| Claude API · Claude Fable 5.1 | $10 / $50 per MTok | List price. Source | 2026-10-06 |
| OpenAI API · GPT-6 Astra | $10 / $50 per MTok | List price. Source | 2026-10-06 |
Corrections
Data lives in a Turso database. Changes go live within seconds, with no redeploy.