AI Coding Field Notes
Coding agent pricing comparison: cheapest LLM API for coding, cost per million tokens, from the OpenRouter price list as JSON and CSV.
AI 编程工具到底花多少钱:335 行数,每行带原句和日期
English · 中文 · Español · 日本語 · 한국어 · Tiếng Việt · Français · Deutsch · Русский · Bahasa Indonesia
这是一份公开数表。32 篇实测笔记里出现过的每一个数 —— 价格、百分比、倍数、token 数、耗时 —— 都抽成一行,每行都附上它原来所在的那句话和发布日期。
跑 agent 的模型,今天多少钱
60 个在 Design Arena 的 agents 各档里排过名的模型,每百万 token 的挂牌价——不是账单:缓存命中、batch、不同接入方各有各的价。取自 OpenRouter 的公开目录,最后核到 2026-08-22。最便宜的三个:
| $ in / 1M | $ out / 1M | 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 |
先看数
下面几行是英文原文照抄,没有翻译:一个数离开它原来那句话就没法核对 —— $1.43 可能是每百万 token、每月、也可能是每个席位。
| 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 |
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
字段里的 published 是那篇笔记发出去的日子,不是这个价钱当时还作数的日子。价格一直在变,每一行都得按它自己的日期读。
正文
正文是英文,在这儿:https://xyzs996.github.io/llm-api-pricing/ —— 带数表、题目页和厂商页。只想要数的话,上面两条 curl 就够了,不用读正文。
说一句
- 收藏这个仓库,数表更新会跟着走。数据是 CC BY 的,收不收藏都随便拿,这儿不拿数据换 star。
- 数不对? 哪个价钱变了、你自己跑出来是另一个数 —— 提个issue。这个仓库就是干这个的。 (issue)
- 想要的数这儿没有? 说清楚要哪个指标、哪家、什么单位,一行就够 —— 表单只有一格必填,收到就收成新行。 (form)
CC BY 4.0:随便复制、转载、再加工、拿去卖。唯一的条件是说一声出处,挂个链接回 https://xyzs996.github.io/llm-api-pricing/ 就行。