作者： Jakub Rusinowski · 最后更新： 2026年6月15日
Free LLM API tiers for prototyping, rented GPUs for fine-tuning and try-before-you-buy, hosted open models when you want Llama-class output without owning a card — the honest map of cloud AI options, from the site that usually tells you to run it locally.
All links are plain links — no referral codes. Entries verified 2026-07-08; tiers change often, confirm on the provider's page.
Genuinely free API access — rate-limited, but real. Enough for prototyping, learning, and low-volume tools before you spend a cent on hardware or tokens.
| Provider | Pricing / free tier | Best for |
|---|---|---|
| NVIDIA NIM (build.nvidia.com) | Free: Free hosted endpoints for Llama, Nemotron, Mistral & more with generous dev credits — sign-up only, no card Free API credits for development; self-host NIM containers with NVIDIA AI Enterprise in production | Trying big open models (Llama 3.3 70B, Nemotron) over an OpenAI-compatible API before buying hardware |
| Google AI Studio | Free: Gemini Flash free tier — the most generous standing free allowance of the big three Free-of-charge tier with per-minute/per-day rate limits; paid tier unlocks higher limits | Prototyping against a frontier-family model at zero cost |
| Groq | Free: Free developer tier with daily rate limits on open models (Llama, Qwen, Gemma) Pay-per-token on LPU hardware; famous for extreme tokens/sec | Latency-critical demos — hundreds of tokens/sec on 70B-class open models |
| OpenRouter | Free: Rotating set of :free model variants with daily request caps One API key, 300+ models, pass-through pricing per model | Comparing many models through one OpenAI-compatible endpoint before committing |
| Mistral La Plateforme | Free: Free experiment tier with rate limits Per-token pricing on Mistral models; EU-hosted | EU-hosted API access with a free on-ramp — relevant for GDPR-sensitive prototyping |
| Cerebras Inference | Free: Free developer tier with daily token caps Per-token pricing on wafer-scale hardware; very high tokens/sec | Fastest hosted open-model inference alongside Groq — good for agent loops |
| GitHub Models | Free: Free playground + API rate limits for GPT-4o, Llama, Phi, Mistral with any GitHub account Included with a GitHub account; production use routes to Azure AI | Zero-signup-friction experimentation if you already live on GitHub |
| Cloudflare Workers AI | Free: Daily free allocation of neurons on every Cloudflare account Per-neuron (unit) pricing beyond the free allocation; runs on Cloudflare edge GPUs | Sprinkling small-model inference (Llama 8B class) into edge apps without managing servers |
| Hugging Face Inference Providers | Free: Small monthly free credit allowance on every HF account Routes to partner providers at pass-through prices; monthly included credits on free & PRO plans | Calling almost any open-weight model on the Hub without picking a vendor first |
Rent a GPU by the hour instead of buying one. The right answer for fine-tuning runs, batch jobs, and "try a 70B before committing to a $2,000 build".
| Provider | Pricing / free tier | Best for |
|---|---|---|
| RunPod | RTX 4090 ≈ $0.3–0.7/hr, A100/H100 by the hour; serverless per-second billing available | Renting a 24–80 GB GPU for hours-long jobs: fine-tuning, batch inference, trying 70B before buying |
| Vast.ai | Marketplace of hosted GPUs — often the cheapest per hour (interruptible options) | Cheapest possible GPU-hours when you can tolerate marketplace variability |
| Lambda | On-demand A100/H100/B200 instances and clusters, per-hour billing | Serious training/fine-tuning runs on datacenter GPUs with clean images |
| Paperspace (DigitalOcean) | Free: Free-tier Gradient notebooks with entry GPUs, subject to availability Per-hour GPU machines + Gradient notebooks; some free-GPU notebook availability | Notebook-first experimentation with an easy upgrade path |
| Hyperstack | Aggressively priced A100/H100 per-hour instances in EU/NA regions | EU-region datacenter GPUs at low hourly rates |
| TensorDock | Marketplace pricing; consumer + datacenter GPUs from ~$0.2/hr | Budget hourly GPUs with per-minute billing |
| Google Colab | Free: Free T4 GPU notebook sessions — the classic zero-cost on-ramp Free notebooks with T4 GPUs (availability-based); Pro tiers add better GPUs and runtimes | Learning, small fine-tunes (LoRA on 7–8B), and quick experiments without any setup |
Open-weight models (Llama, Qwen, Mistral, DeepSeek) served per-token by specialists — the middle ground between frontier APIs and running your own hardware.
| Provider | Pricing / free tier | Best for |
|---|---|---|
| Together AI | Free: Small starter credit for new accounts Per-token pricing on 100+ open models (Llama 3.1 8B ≈ $0.18/M tokens) | Production open-model serving with serious throughput and fine-tuning endpoints |
| Fireworks AI | Free: Starter credit for new accounts Per-token open-model serving; fast function-calling-optimized deployments | Low-latency production inference on open models, incl. dedicated deployments |
| DeepInfra | Among the lowest per-token prices for open models; OpenAI-compatible | Cheapest hosted per-token rates on Llama/Qwen/Mistral-class models |
| Replicate | Per-second (hardware-based) or per-token pricing; huge community model catalog | Running community models (image, audio, video, LLM) without packaging them yourself |
| Novita AI | Free: Small trial credit Budget per-token open-model APIs + GPU instances | Low-cost open-model APIs with an Asia-Pacific footprint |
The closed frontier models. Highest quality, zero setup, per-token billing — the benchmark every local setup gets compared against.
| Provider | Pricing / free tier | Best for |
|---|---|---|
| OpenAI API | GPT-4o $2.50/M input, $10/M output; GPT-4o Mini $0.15/$0.60 | Frontier quality and the broadest tooling ecosystem |
| Anthropic API | Claude Sonnet $3/M input, $15/M output; Haiku $0.80/$4 | Best-in-class coding and long-context analysis |
| Google Gemini API | Free: Free tier via Google AI Studio (see above) Gemini Flash $0.10/M input, $0.40/M output; Pro $1.25/$5 | Huge context windows and the lowest frontier-family prices |
Twelve editorial comparisons with a shared break-even calculator. Start with the flagship: Local AI vs Cloud API Costs: The Real Break-Even Math (2026).
Local vs cloud cost calculator | Local AI build guide | GPU benchmarks