发现各大提供商提供的免费大模型 API,对比速度与延迟数据。免费模型 2772服务商 157免费服务 8807
| 模型 | 服务商 | 速度 | 延迟 | 测试 |
|---|---|---|---|---|
claude-3-haiku-20240307Anthropic免费 文本 | +1 更多收起 | N/A | N/A | 0 |
claude-3-haiku-20241022Claude免费 文本 | N/A | N/A | 0 | |
LMSpeed 追踪了来自 157 家 API 提供商的 2772 个免费大语言模型。各提供商的免费额度不同——有的限制每日请求数,有的为新用户提供免费额度。所有速度数据均来自真实 API 测试。
先找到合适的免费模型,对比它背后的服务商,再进入模型或服务商详情页确认限制与表现。
使用搜索、模型系列筛选和能力标签,把目录缩小到真正想测试的免费模型。
查看服务商数量、测试量、吞吐速度和首字延迟,不只按价格选择。
在服务商详情页确认可用性、健康检测、价格备注和免费额度限制,再用于测试或集成。
关于免费额度与如何使用这份目录的常见问题。
| N/A |
| N/A |
| 0 |
claude-sonnet-4-6-thinkClaude免费 文本 | N/A | N/A | 0 |
claude-3-opus-20240229Anthropic免费 ToolsFilesVision200K | N/A | N/A | 0 |
claude-3-opus-latestClaude免费 文本 | N/A | N/A | 0 |
codegemma-1.1-7b免费 | N/A | N/A | 0 |
xopdeepseekv32DeepSeek免费 | N/A | N/A | 0 |
xopglm51免费 | N/A | N/A | 0 |
nvidia/nemotron-nano-12b-v2-vl:free免费 | N/A | N/A | 0 |
google/gemma-4-26b-a4b-it:free免费 | N/A | N/A | 0 |
nvidia/nemotron-nano-9b-v2:free免费 | N/A | N/A | 0 |
nvidia/nemotron-3-nano-30b-a3b:free免费 | N/A | N/A | 0 |
nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free免费 | N/A | N/A | 0 |
nvidia/nemotron-3-super-120b-a12b:free免费 | N/A | N/A | 0 |
google/gemma-4-31b-it:free免费 | N/A | N/A | 0 |
cognitivecomputations/dolphin-mistral-24b-venice-edition:freeMistral免费 | N/A | N/A | 0 |
meta-llama/llama-3.2-3b-instruct:freeMeta免费 | N/A | N/A | 0 |
meta-llama/llama-3.3-70b-instruct:freeMeta免费 | N/A | N/A | 0 |
qwen/qwen3-next-80b-a3b-instruct:free阿里巴巴免费 | N/A | N/A | 0 |
poolside/laguna-m.1:free免费 | N/A | N/A | 0 |
liquid/lfm-2.5-1.2b-thinking:free免费 | N/A | N/A | 0 |
poolside/laguna-xs.2:free免费 | N/A | N/A | 0 |
gemini-3.1-flash-tts-previewGoogle免费 | N/A | N/A | 0 |
cnb免费 | N/A | N/A | 0 |
GPT-5.6 Luna is a fast, cost-efficient model in OpenAI's GPT-5.6 series. It is suited for high-volume, latency-sensitive tasks such as chat, classification, and lightweight agentic workflows, providing capable reasoning for... | N/A | N/A | 0 |
GPT-5.6 Terra is a balanced model in OpenAI's GPT-5.6 series, positioned between the flagship Sol tier and the cost-efficient Luna tier. It is suited for everyday coding, reasoning, and agentic... | N/A | N/A | 0 |
GPT-5.6 Sol is the flagship model in OpenAI's GPT-5.6 series. It is suited for complex reasoning, coding, and agentic workflows, and is particularly strong at command-line and multi-step coding tasks... | N/A | N/A | 0 |
NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it... | 25.13 t/s | 8.76 s | 5 |
Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort:... | N/A | N/A | 0 |
NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications. Built on a hybrid Mamba-Transformer... | N/A | N/A | 0 |
Trinity Large Thinking is a powerful open source reasoning model from the team at Arcee AI. It shows strong performance in PinchBench, agentic workloads, and reasoning tasks. Launch video: https://youtu.be/Gc82AXLa0Rg?si=4RLn6WBz33qT--B7... | N/A | N/A | 0 |
NVIDIA Nemotron™ 3 Nano Omni is a 30B-A3B open multimodal model designed to function as a perception and context sub-agent in enterprise agent systems. It accepts text, image, video, and... | N/A | N/A | 0 |
ReasoningToolsFilesOpen WeightsMiniMax M3 is MiniMax next-generation large language model, designed for advanced reasoning, long-context understanding, and high-quality multilingual dialogue. | N/A | N/A | 0 |
ReasoningToolsFilesVisionAnthropic 的 Claude Opus 4.7是旗舰大语言模型,专为高复杂度通用任务与多步推理优化。 | 兔子API我不是AI神Moyanjdc API +39 更多收起Higobs APIKternaDMXAPICodeXEPrivnode丰思理 AISkyAIVenlacyCan API17NAS API情酱的API站天宫造物丸美小沐写作丸美小沐MagicAI艾可APIQWQ Chat APIAIO通用智能服务平台Fangyuan APILiunew APIAIsa42公益站DeepRouterKoyeb AI GatewaySWT-APIZetaTechs APISwifllyLLM毫秒APIMentoe APIDibin84 API HubNuizi APIRinkoAIJuCodeHank Workspace APISoul 公益站逆龙傲公益站AI APIWAADRIMyNav AI | N/A | N/A | 0 |
ToolsVision131.1KLlama 4 Maverick 17B Instruct (128E) is a high-capacity multimodal language model from Meta, built on a mixture-of-experts (MoE) architecture with 128 experts and 17 billion active parameters per forward... | N/A | N/A | 0 |
Tools131.1K英伟达Qwen3-Next-80B-A3B-Instruct is an instruction-tuned chat model in the Qwen3-Next series optimized for fast, stable responses without “thinking” traces. It targets complex tasks across reasoning, code generation, knowledge QA, and multilingual... | N/A | N/A | 0 |
mistral-small-2603Mistral免费 ReasoningToolsFilesOpen Weights | 57.26 t/s | 3.90 s | 10 |
gemini-3.5-flashGoogle免费 | 231.45 t/s | 3.64 s | 5 |
step-3.7-flash免费 | 71.74 t/s | 17.60 s | 5 |
cerebras/zai-glm-4.7智谱免费 | 395.00 t/s | 2.48 s | 5 |
c4ai-aya-expanse-32b免费 | N/A | N/A | 0 |
gpt-image-2-psdOpenAI免费 | N/A | N/A | 0 |
openai-5.4OpenAI免费 | N/A | N/A | 0 |
openai-5.4-xhighOpenAI免费 | N/A | N/A | 0 |
gpt-image-2-pptOpenAI免费 | N/A | N/A | 0 |
openai-5.5OpenAI免费 | N/A | N/A | 0 |
bge-small-zh-v1.5免费 | N/A | N/A | 0 |
openai-5.5(xhigh)OpenAI免费 | N/A | N/A | 0 |
bce-reranker-base_v1免费 | N/A | N/A | 0 |
bce-embedding-base_v1免费 | N/A | N/A | 0 |
hy3-free免费 | N/A | N/A | 0 |
openai-5.5-xhighOpenAI免费 | N/A | N/A | 0 |
dolphin-mistral-24b-venice-editionMistral免费 | N/A | N/A | 0 |
diffusiongemma-26b-a4b-it免费 | N/A | N/A | 0 |
all-mpnet-base-v2免费 | N/A | N/A | 0 |
doubao-1-5-pro字节跳动免费 | N/A | N/A | 0 |
nvidia/nemotron-3.5-content-safety:free免费 | N/A | N/A | 0 |
jina-clip-v1Jina免费 | N/A | N/A | 0 |
command-r7b-12-2024Cohere免费 | N/A | N/A | 0 |
指的是某家提供商上输入和输出价格都为 $0 / token 的模型组合。部分提供商提供长期免费额度,部分仅向新账号一次性赠送。只有在公开价格为零时才会被标记为免费,价格页会定期复核。
大多数有速率限制——每分钟请求数、每日总量、上下文长度上限——很多要求实名注册并绑定手机号或支付方式。部分是阶段性活动。在依赖某个免费端点之前,请仔细阅读提供商的条款和配额。
部分条目是社区运营的「公益站」或聚合中转——站长用自掏腰包的上游 key 打包再免费分发。它们通常额度更大、可用模型更全,但稳定性远低于官方 API:站长随时可能停服或「跑路」,额度和价格也可能毫无预警地变动;很多站点是邀请制,需要 GitHub 邀请码、社区推荐或特定圈子才能注册,部分长期关闭注册。建议把它们当作可遇不可求的备用通道,重要工作流仍走官方付费端点。
速度因模型和提供商而异。按「最多测试」排序找数据最稳的,或在上方按模型系列筛选。每行展示真实 API 测试得到的 tokens/秒 中位值与首字延迟。
对每家提供商使用相同 prompt 做五轮压力测试,使用 tiktoken 统计输出 token 数,测量吞吐量(tokens/秒)和首字到达时间。所有指标以中位数聚合以抵抗异常值,并按固定节奏刷新。
用于原型、个人项目和低流量工具是可以的。生产流量很快就会撞到速率限制。建议把免费额度当成「试用通道」:先验证模型和服务商表现,再切到同模型的付费端点上线。
可能目前没有提供商提供免费版本、免费活动已结束,或者尚未被基准测试覆盖。可以在该模型的详情页查看付费方案,或通过页脚反馈链接告诉我们漏了哪家。