76 model releases in a year — and Flash/Haiku tiers cost up to 120× less per token than flagships

Asked:

dates of model releases from google, anthropic, openai in the last year, with corresponding costs per 1m tokens

A scan of first-party Google, Anthropic and OpenAI pages found 76 release-labelled model or version rows in the rolling year 2025-09-03 to 2026-09-03 — 35 OpenAI, 28 Google, 13 Anthropic. For 29 of them a standard text-token list price (USD per 1M tokens, captured 2026-09-03) matched on the official pricing pages; the rest are product-only, preview, ChatGPT-only, specialized or snapshot announcements without a directly matched API price and are shown blank, never as zero.

Google

releases · 9 with matched input price

Anthropic

releases · 11 with matched input price

OpenAI

releases · 9 with matched input price

Release timeline — a year of launches, priced where a price matched

Each dot is one release, placed by announcement date. Filled dot = matched output price (dot size scales with USD per 1M output tokens); hollow dot = no matched standard text-token price. Hover any dot for details.

Price spread — input vs output, log scale

The 28 releases with both input and output prices matched. Both axes log-scaled, USD per 1M tokens. Cheap Flash/Lite/Haiku tiers sit bottom-left; premium reasoning flagships top-right.

Findings

Cheapest matched release of the year is Google's Gemini 3.1 Flash-Lite ($0.25 in / $1.50 out per 1M), versus $30 in / $180 out for OpenAI's GPT-5.4 Pro and GPT-5.5 Pro — a 120× gap on both sides. ai.google.dev · developers.openai.com

Anthropic held its tier prices remarkably steady: Opus 4.5 through Opus 5 all list $5 in / $25 out, and Sonnet 4.5 and 4.6 both list $3 in / $15 out. docs.anthropic.com

Cache pricing is where the deepest discounts live: OpenAI's GPT-5.6 Luna lists cached input at $0.02 per 1M — 50× below its $1.00 fresh-input rate. developers.openai.com

OpenAI announced the most rows (35) but only 9 matched a standard API token price — most GPT-5.x Instant/Thinking/Pro rows are ChatGPT-product releases, not separately billed API SKUs. help.openai.com

Every release — sortable

Click a header to sort. Blank price = no matched standard text-token list price on the collected official pages ("not matched"), not zero. Cached input is cache-read pricing for Anthropic, cached-input for OpenAI; Google prices use the up-to-200k-token tier where tiered.

Date	Company	Model	Input $/1M	Cached in $/1M	Output $/1M	Release source	Pricing

Method: 76 release-labelled model/version rows scraped from first-party Google, Anthropic and OpenAI release notes, changelogs and announcement pages, covering 2025-09-03 to 2026-09-03. Prices are standard text list prices in USD per 1M tokens from official pricing pages captured 2026-09-03 and matched for 29 rows; dates are first public announcement/release/preview dates for the exact named row. Prices can change after launch. Blank means no matched price, never zero. This is a broad release log, not a claim that every row is a separately billed API SKU.

This report was generated automatically by Keenable SELECT at a user's request, from publicly available web sources linked herein. Keenable does not review, verify, or endorse its contents and makes no representation as to accuracy, completeness, or timeliness; AI-based extraction may contain errors. Nothing in this report is investment, legal, financial, or other professional advice. All trademarks and referenced content remain the property of their respective owners; no affiliation or endorsement is implied. To report an error, rights concern, or request removal: legal@keenable.ai.