本文来源:http://www.lpqz.cn/news/167615
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
emini、NVIDIA NIM、Groq、OpenRouter 等主流免费大模型API的上下文窗口与速率限制做成一张表帮你快速选出最适合写代码、长文档或 Agent 场景的那一个。数据每日自动刷新最近更新2026-09-24不怕信息过期。一、什么是 awesome-freellm-apis一份会保鲜的免费大模型API目录大多数免费 API 合集是静态清单——写好后一两年没人更新等你照着申请免费额度早就没了。这个项目做对了几件关键的事亮点说明✅ 数据每日更新通过自动化监控每天刷新各家免费额度而非过期的静态列表✅ 信用卡透明度每个供应商都标注是否需要信用卡、手机验证还是啥都不要✅ 一键配置片段内置 Claude Code、Cursor、Codex 等 10 工具的即用配置✅ 横向对比上下文窗口、速率限制、模态支持一张表并排比较完整目录、Base URL 速查表和模型清单都写在 README.
md 中另有简体中文版 README.zh-CN.md、繁体、日文、韩文版随取随用。二、看懂对比表先认识 4 个关键指标在横向对比前花 1 分钟搞懂术语后面看表就不费劲了 上下文窗口Context Window模型一次能记住并处理多少 token。128K≈10 万中文字1M≈70 万字。做长文档问答、喂整个代码库时这个数字越大越从容。RPMRequests Per Minute每分钟最多请求数决定短时间的连发能力。RPDRequests Per Day每天最多请求数决定长期免费额度。TPM / TPDTokens Per Minute / Day按 token 量计的限速长 prompt 场景更吃这个。小建议RPM 看顺不顺RPD 看够不够。两者要结合着看。三、上下文窗口横向对比谁的记性最长以下摘自 README.md 的供应商目录按最大上下文窗口降序精选供应商最大上下文窗口免费模型数需
README.md 最佳模型表供应商免费层速率限制需要信用卡适合谁Groq30 RPM14,400 RPD不需要免费额度最慷慨首选入门Google Gemini15 RPM1,500 RPD不需要1M 上下文 多模态全能型Mistral AI约 1 RPS500K TPM不需要Medium 3.5 支持 256KCohere20 RPM不需要长文本 / RAG 场景强Kilo Code200 请求/小时不需要Nemotron 3 Ultra 550B 免费SambaNova20 RPM20 RPD200K TPD注册即可推理速度快Cerebras10 RPM100 RPD1M TPD不需要芯片级低延迟LLM7.io10 RPM60 请求/小时不需要匿名也能用AI21 Labs200 RPM10 RPS注册即可每分钟吞吐最高Z AI智谱1 个并发请求不需要GLM-4.7-Flash 免费体
、Cerebras 的低延迟端点或用OpenRouter免费模型做统一路由一个 Key 调多家模型。现成的配置片段见 code-examples/cursor.md、code-examples/claude-code.md、code-examples/codex.md。六、3 步把免费API接入工具从 Key 到跑通选供应商并注册——对照上面两张表选一个不需要信用卡的邮箱注册通常 1 分钟内拿到 Key。复制 Base URL 模型 ID——速查表在 README.md 的 Quick Reference 一节所有供应商均为 OpenAI 兼容端点例如 Groq 的端点地址为https://api.groq.com/openai/v1。粘贴进你的工具——Cursor 在 Settings → Models → Add Model 中填入 Key 与 Base URLClaude Code
/ Codex 通过环境变量配置。具体步骤直接抄 code-examples/ 目录下的对应文档即可。遇到 rate limit429错误说明触顶了——等下个窗口刷新或换一家供应商继续这正是多备几个 Key 的意义。七、数据来源与项目结构 数据每日自动刷新最近更新时间2026-09-24对比的是今天的真实免费额度。项目结构精简核心文件文件内容README.md完整供应商目录、Base URL 速查、最佳免费模型表README.zh-CN.md简体中文版code-examples/cursor.md 等 3 个Cursor / Claude Code / Codex 即用配置CONTRIBUTING.md收录标准与贡献指南八、常见问题 FAQQ1免费额度是永久的还是试用赠送两者都有且表里分得很清楚主体是永久免费层Permanent Free Tiers另有 OpenRouter 等可续充额
度单独列出。收录标准见 CONTRIBUTING.md纯广告、无真实免费额度的服务不会入选。Q2上下文窗口越大越好吗不一定。窗口大意味着单次请求 token 成本高、免费额度消耗快。128K–262K 对绝大多数编程与对话场景足够1M 留给一次塞进整本书的极端需求。Q3一定要用信用卡吗不是。对比表中Groq、Google Gemini、Mistral、Cohere、Cloudflare、Cerebras等主流选项均标注不需要邮箱注册即可上手。总结入门首选Groq免费额度最慷慨长文档选NVIDIA NIM / Gemini / ModelScope1M 窗口编程助手用Groq OpenRouter组合最稳。所有对比数据每日更新收藏 README.md 这一份活的免费大模型API目录就能长期用上新鲜准确的额度信息 【免费下载链接】awesome-freellm-apis134 free LL
M APIs AI API keys from 40 providers. Google Gemini, NVIDIA NIM, Groq, OpenRouter more. One-click setup for Claude Code, Cursor and Codex.项目地址: https://gitcode.com/gh_mirrors/aw/awesome-freellm-apis创作声明:本文部分内容由AI辅助生成(AIGC),仅供参考
本文来源:http://www.lpqz.cn/news/167615
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Which free large model API is the best? awesome-freellm-apis Context window and rate limit side-by-side comparison guide [Free download link] awesome-freellm-apis 134 free LLM APIs AI API keys from 40 providers. Google Gemini, NVIDIA NIM, Groq, OpenRouter, and more. One-click setup for Claude Code, Cursor, and Codex.Project URL: https://gitcode.com/gh_mirrors/aw/awesome-freellm-apis free large model API is the best? How large is the context window? How many free calls can be made per day? This article is based on the open-source project awesome-freellm-apis, which compiles 508 free LLM APIs and API key access points from 31 providers. It includes Google G
The context windows and rate limits of mainstream free large model APIs such as emini, NVIDIA NIM, Groq, and OpenRouter are compiled into a table to help you quickly select the one that is best suited for writing code, long documents, or Agent scenarios. Data is automatically refreshed daily. Last updated on 09/24/2026. No need to worry about outdated information. I. What is awesome-freellm-apis? It is a directory of free large model APIs that stays up-to-date. Most free API collections are static lists — once they're written, no one updates them for a year or two, and by the time you go to apply for the free quota, it's long gone. This project has done several key things right. Highlights: ✅ Data updated daily: Through automated monitoring, the free allowances of each provider are refreshed daily, rather than being an outdated static list. ✅ Credit card transparency: Each provider is marked with whether a credit card, phone verification, or nothing at all is required. ✅ One-click configuration snippets: Built-in ready-to-use configurations for 10 tools, including Claude Code, Cursor, and Codex. ✅ Side-by-side comparison of context windows, rate limits, and modal support: A table provides a side-by-side comparison. The complete directory, Base URL quick reference table, and model list are all included in the README.
The md file also contains a Simplified Chinese version, README.zh-CN.md, as well as Traditional Chinese, Japanese, and Korean versions that can be used as needed. 2. To understand the comparison table, first familiarize yourself with 4 key metrics. Take 1 minute to understand the terminology before making horizontal comparisons. After that, reading the table will be effortless. Context Window: The number of tokens the model can remember and process at once. 128K ≈ 100,000 Chinese characters; 1M ≈ 700,000 characters. When handling Q&A for long documents or feeding an entire codebase, the larger this number, the more comfortable the process. RPM (Requests Per Minute): The maximum number of requests per minute determines the ability to make continuous requests over a short period of time. RPD (Requests Per Day): The maximum number of requests per day determines the long-term free quota. TPM / TPD Tokens Per Minute / Day: The rate limit measured in tokens. Long prompt scenarios are more affected by this. A small suggestion: Check whether the RPM is smooth and whether the RPD is sufficient. Both should be considered together. III. Context Window Comparison: Which Has the Longest Memory? The following is excerpted from the vendor directory in README.md. Selected vendors are listed in descending order of maximum context window. Maximum Context Window Number of Free Models Required
README.md Best Models Table Provider Free Tier Rate Limit Credit Card Required Suitable For Groq30 RPM14,400 RPDNot required Most generous free tier Best for getting started Google Gemini15 RPM1,500 RPDNot required 1M context Multimodal, all-rounder Mistral AIApproximately 1 RPS500K TPMNot required Medium 3.5 support 256K Cohere20 RPMNot required Good for long text / Strong in RAG scenariosKilo Code200 requests/hourNot requiredNemotron 3 Ultra 550B FreeSambaNova20 RPM20 RPD200K TPDRegistration requiredFast inference speedCerebras10 RPM100 RPD1M TPDNot requiredChip-level low latencyLLM7.io10 RPM60 requests/hourNot requiredCan be used anonymouslyAI21 Labs200 RPM10 RPSRegistration requiredHighest throughput per minuteZhiPu AI1 concurrent requestNot requiredGLM-4.7-Flash Free tier
, Cerebras' low-latency endpoint, or use OpenRouter's free model for unified routing to access multiple models with a single key. Ready-to-use configuration snippets can be found in code-examples/cursor.md, code-examples/claude-code.md, and code-examples/codex.md. 6. 3 steps to integrate a free API into a tool, from obtaining the key to getting it up and running: Choose a provider and register – refer to the two tables above to select an email registration that does not require a credit card. You can usually get the key within 1 minute. Copy the Base URL and Model ID – The quick reference table is in the Quick Reference section of README.md. All providers have OpenAI-compatible endpoints. For example, Groq's endpoint address is https://api.groq.com/openai/v1. Paste it into your tool – In Cursor, go to Settings → Models → Add Model and enter the Key and Base URL. Claude Code
/ Codex is configured through environment variables. For specific steps, simply copy the corresponding document under the code-examples/ directory. Encountering a rate limit 429 error means you've hit the limit — wait for the next window to refresh or switch to another provider to continue. This is precisely why it's important to have several keys on hand. 7. Data Sources and Project Structure Data is automatically refreshed daily. The most recent update was on 09-24-2026. The comparison is with today's actual free quota. Project Structure Streamlined Core Files File Contents README.md Complete list of providers, quick reference for base URLs, and table of best free models README.zh-CN.md Simplified Chinese version code-examples/cursor.md and 2 other files Ready-to-use configurations for Cursor / Claude Code / Codex CONTRIBUTING.md Inclusion criteria and contribution guidelines VIII. Frequently Asked Questions (FAQ) Q1: Is the free quota permanent or a trial offer? Both are available, and they are clearly distinguished in the table. The main focus is on permanent free tiers. There are also options like OpenRouter that offer top-up credits.
is listed separately. For inclusion criteria, see CONTRIBUTING.md. Services that are purely advertisements and do not offer genuine free usage limits will not be included. Q2: Is a larger context window better? Not necessarily. A large window means high token costs per request and rapid consumption of the free quota. 128K–262K is sufficient for the vast majority of programming and conversational scenarios. 1M is reserved for the extreme requirement of fitting an entire book into a single request. Q3: Is a credit card absolutely necessary? No. In the comparison table, mainstream options such as Groq, Google Gemini, Mistral, Cohere, Cloudflare, and Cerebras are all marked as being available without the need for email registration. Summary: For getting started, Groq is the preferred choice, offering the most generous free quota. For long documents, choose NVIDIA NIM / Gemini / ModelScope with a 1M window. For programming assistants, the combination of Groq and OpenRouter is the most stable. All comparison data is updated daily. Bookmark README.md. This living directory of free large model APIs provides long-term access to up-to-date and accurate quota information. [Free Download Link] awesome-freellm-apis134 free LL
M APIs AI API keys from 40 providers. Google Gemini, NVIDIA NIM, Groq, OpenRouter, and more. One-click setup for Claude Code, Cursor, and Codex.Project URL: https://gitcode.com/gh_mirrors/aw/awesome-freellm-apis: Parts of this article were generated with the assistance of AI (AIGC) and are for reference only.
