LLM API pricing for coding agents, re-read daily

English · 中文 · Español · 日本語 · 한국어 · Tiếng Việt · Français · Deutsch · Русский · Bahasa Indonesia

Two things: a price table re-read from OpenRouter’s catalog every day, and 32 write-ups on what those bills looked like in production.

What the agent models cost (60 models)

List price per million tokens, with the agents rank the Design Arena gave it and the category it ranked in. Last read 2026-08-22. The three cheapest:

$ in	$ out	Model	Best agents rank
$0.1875	$0.9375	Gemini 3.7 Flash `batch`	#3 androidnative
$0.25	$1.50	Gemini 3 Flash Preview `batch`	#9 agenticslides
$0.30	$1.20	MiniMax M3	#10 python-pptxslides

All 60 models · JSON · CSV

List price, not your bill — cache hits, batch tiers and providers price differently, and batch rows are kept separate rather than folded in. The prices are the catalog’s and the ranks are the arena’s; neither is ours, and both move.

Star this repository to bookmark the table and follow releases. The data is CC BY: a star changes nothing about what you can do with it, and this project gives nothing in return for one.

Read these on the web: AI Coding Field Notes — the same write-ups with the figures table, the topic pages and the links between them.

On the figures. Every number here traces back to a named source or a run we did. Numbers we could not trace were cut before publishing, not rounded or guessed. Each piece says up front that it was drafted with AI assistance.

The figures, as data (335 rows)

Every figure published across the 32 write-ups — prices, percentages, multiples, token counts and durations — pulled into one table. Each row carries the full sentence it came from and a link to the piece, so you can check it without reading all of them.

One row from each of the 12 most recent write-ups — quoted verbatim, not summarised:

Figure	The sentence it came from	Write-up
`$1.43`	The $1.43 and the $9.05 are both frontier models doing a job they were not specifically built for.	The Two Best AI Code Reviewers Score the Same. One Costs $1.43 a Run, the Other $9.05.
`$29`	Before writing a contract-comparison tool, one builder handled three to ten comparisons by hand at $29 a document, and only turned the routine into software once the same people kept coming back and paying for it.	Debunking the Myth of Overnight Success in Micro-SaaS
`90%`	90% of developers still rely on manual prompt writing, while top performers use Skill Package to automate 80% of repetitive tasks, saving hours weekly.	Best Practices for AI Agent Skill Management
`80%`	When the Claude Code team decided to slash 80% of their system prompts, most developers expected the model to lose its edge in complex engineering tasks.	Why Stripping 80% of System Prompts Actually Improved Claude Code’s Performance
`$30`	With a budget as low as $30 per day, developers have reached an effective lead cost of $3 to $4.	Stop Reading SimilarWeb Like a Traffic Dashboard — Read It Like a Feasibility Test
`$1,000`	Instead of chasing a 2.5% consumer conversion rate across unpredictable social channels, you sell a single $1,000 to $5,000 service package directly to one business owner — no massive ad campaigns, no hundreds of low-tier support tickets.	How Indie Developers Are Building AI-Powered “Digital Landlords” and Renting Them Out for Monthly Cash Flow
`40-second`	When an independent developer uses Agency Agents to set up a 40-second response cycle for e-commerce listings, they are building a feedback loop that reads market conditions and adjusts, which is what separates a timed automation from a script on a timer.	Stop Chatting With AI: How I Use /loop and /hook to Automate My Indie Dev Workflow
`9.1%`	One reported case moved entry-group conversion from 9.1% to 55.1% by rebuilding an automated onboarding flow around what the funnel data actually showed, rather than around what the team assumed users were doing.	Boosting AI Bot Conversion: A Deep Dive into Funnel Data
`$1.25`	Meta priced Muse Spark 1.1 at $1.25 per million input and $4.25 per million output, roughly 75% and 83% below Anthropic’s Opus, and the tradeoff is visible in the benchmarks, since it leads on MCP Atlas and JobBench while trailing on SWE-Bench Pro and DeepSWE 1.1.	1.6 Billion Free Tokens Is a Compression Ratio, Not a Strategy
`9x`	Open Code Review reports roughly 9x lower token consumption than general-purpose agents while holding accuracy, which suggests that a specialized agent aimed at one job often beats a heavy generalist on the only axis an indie developer can afford to optimize.	Token Optimization for Indie Developers’ AI API Bills
`87 percent`	The 87 percent figure and those two cases are measuring the same thing from different angles, which is time spent moving data by hand between a system that knows the answer and a system that needs it.	The 5 AI Features That Separated 27 Profitable Solopreneurs From the Rest
`40 seconds`	By using an AI agent to automate responses, negotiation, and price adjustment, the average response time can be reduced to within 40 seconds.	From AI Demo to Product: Loop Engineering for Indie Devs

All 335 rows — or as data:

curl -s https://cdn.jsdelivr.net/gh/xyzs996/llm-api-pricing@main/data/figures.json
curl -s https://cdn.jsdelivr.net/gh/xyzs996/llm-api-pricing@main/data/figures.csv

Fields: value, kind (price / percent / multiple / tokens / duration), unit, context, article, published, url, medium. published is the day that write-up went out, not the day the figure was in force — prices move, so read each row as of its own date. Browse it as a table: figures.md.

Those two go through jsDelivr, which caches @main for up to 12 hours — fine for a table that is rebuilt once a day. If you want it uncached, the origin is https://xyzs996.github.io/llm-api-pricing/data/figures.json.

A number without its sentence is not checkable — $1.43 could be per million tokens, per month, or per seat. The sentence is quoted verbatim, not summarised.

Citing a row? @main moves — today’s row 88 may be somewhere else tomorrow. Every snapshot is also frozen as a dated release, and these two always resolve to the newest frozen one:

curl -sL https://github.com/xyzs996/llm-api-pricing/releases/latest/download/figures.json
curl -sL https://github.com/xyzs996/llm-api-pricing/releases/latest/download/figures.csv

All snapshots — one per day the table actually changed.

Corrections wanted. If a figure is stale, a tool changed its pricing, or you ran the same thing and got something else — open an issue. That is what this repo is for.

Want a figure that is not here yet? Say which metric, which provider, which unit — in one line. The form has one required field and that is the whole of it.

Or read one of these first. Each answers a question people actually search for, with every figure this repo has on it — the sentence and the date included:

AI agent loop engineering: what does the 5x claim actually measure? — every figure published about running an agent in a loop instead of re-prompting it — including the one the write-up itself refuses.
Which Chinese AI agent tools are actually free for coding, and what do they cost once they are not? — the per-million input prices behind the free-token grants, each kept with the sentence and the date it was published in.
Which AI programming tool should I pick — and does the 70% time-saving figure apply to me? — why that one widely-quoted time-saving figure holds for one kind of reader and is close to meaningless for the other.
Two AI code reviewers score within two points. One costs a sixth as much per run. — both benchmark scores and both per-run prices side by side, with the vendor self-reports marked as such.
What does a coding agent actually cost per month — and which figure on the pricing page is the one that moves it? — every published per-million price this repo has collected, and why the one that decides the bill is usually not on the page.
Klarna saved $4M replacing 700 agents, then rehired. Which number arrived first, and which one would have warned you? — the savings, the satisfaction drop and the months between them, in the order they were published.
What does a one-person software product actually make — and how many months before it made anything? — the published monthly revenue figures next to the months each one took to get there, including the eighteen-month one.
Short video for indie products: 95% organic sounds great until you ask how many hours a week it costs — the reach figures next to the weekly hour counts behind them, and why an organic share with no denominator is not a result.

Follow along. Atom feed — new write-ups land there first.

Reading this with a model? llms.txt — the dataset first, then every write-up with one line of what it says.

By provider. GPT-5.6 (18) · Klarna (9) · BrowserAct (7) · Claude (7) · WorkBuddy (7) · Fable 5 (6) · ChatGPT (4) — every figure whose sentence names it, with the date.

By topic. Indie Development (12) · Automation Systems (10) · SaaS Business (7) · AI Costs (6) · AI Implementation (6) · Niche Market (6) · AI Programming (5) · Cost Savings (5) · Productivity (5) · Development Tools (4) · AI Features (3) · AI Tools (3) · Artificial Intelligence (3) · Chinese AI (3) · Code Review (3) · Micro SaaS (3) · Token Optimization (3)

free-llm-api — verified free LLM API tiers, rate limits, and no-card options.

Reuse

One condition: say where it came from. A link back to https://xyzs996.github.io/llm-api-pricing/ next to whatever you reuse is enough.

AI Coding Field Notes

LLM API pricing for coding agents, re-read daily

What the agent models cost (60 models)

The figures, as data (335 rows)

The write-ups

How Chinese AI Agent Tools Leverage 1.6 Billion Free Tokens

The Cost-Effective Guide to Using Open Code Review for AI Programming Tools

1.6 Billion Free Tokens Is a Compression Ratio, Not a Strategy

58 Million Plays Started With One Account, Not Four

A 30-Line Script, 200 Users, and a Niche Nobody Wanted

AI Agent Loop Engineering: Karpathy’s Method for 5x Productivity Gains

AI Programming Tool Selection Strategy: From Rapid Prototyping to Long-term Collaboration

AI-Generated Local Business Websites Don’t Rent for $3,000/Month (Until You Do This)

Best Practices for AI Agent Skill Management

Beyond Chat: How Codex Can Automate Your Word/Excel/PPT/PDF Workflows

Beyond Token Pricing: How Indie Devs Should Really Evaluate AI Model Costs

Boosting AI Bot Conversion: A Deep Dive into Funnel Data

Choosing the Right AI Model for Coding: Cost vs. Efficiency

Debunking the Myth of Overnight Success in Micro-SaaS

From AI Demo to Product: Loop Engineering for Indie Devs

How Chinese Developers Are Using Codex Record & Replay to Streamline Repetitive Workflows

How Indie Developers Are Building AI-Powered “Digital Landlords” and Renting Them Out for Monthly Cash Flow

How to Build a Micro-SaaS Without Spending a Dime on Ads

MonkeyCode: The Open-Source AI Coding Platform With 900 Million Free Tokens

Stop Chatting With AI: How I Use /loop and /hook to Automate My Indie Dev Workflow

Stop Reading SimilarWeb Like a Traffic Dashboard — Read It Like a Feasibility Test

The 5 AI Features That Separated 27 Profitable Solopreneurs From the Rest

The AI Branding Revolution: How Indie Developers Are Ditching Design Costs with AI

The First Line of Defense in AI Programming: Environment Variable Management

The Hidden Costs of GPT-5.6 Model Selection: A Developer’s Real-World Guide

The Klarna Lesson: Why AI Implementation Needs a Staircase, Not a Leap

The Two Best AI Code Reviewers Score the Same. One Costs $1.43 a Run, the Other $9.05.

Token Optimization for Indie Developers’ AI API Bills

When AI Customer Service Backfired: Klarna’s Case and the Four-Stage Path to Enterprise AI Adoption

Why Pi’s 1000-Token Agent Engine Needs a Sandbox Before You Touch It

Why Stripping 80% of System Prompts Actually Improved Claude Code’s Performance

Why Your Indie App Needs Short-Form Video Marketing (And How to Get Started)

Reuse

LLM API pricing for coding agents, re-read daily

What the agent models cost (60 models)

The figures, as data (335 rows)

The write-ups

Related

Reuse