Free tier: AI tools and services

Free tier: AI tools and services

Free tiers for calling models, and for the less visible work around them: evaluating output, tracing what a model actually did, and running notebooks. The evaluation and observability tools are the larger share, which reflects where the practical difficulty sits.

Part of the Free Tier Services Directory.

Running and calling models

• Gonka Broker — OpenAI-compatible API serving open-source models over a decentralised GPU network.
• Lumenfall.ai — Unified API for several image generation models, using the OpenAI request format.
• Mediaworkbench.ai — Access to Azure OpenAI, DeepSeek and Gemini models. 100,000 words free.
• OpenRouter — Single API in front of many models, including free access to DeepSeek R1 and V3, Llama and Moonshot.
• Pollinations.AI — Image generation with a free API, no signup and no key required.
• Portkey — Gateway and control panel for generative AI applications, combining routing across providers with logging. 10,000 requests a month.
• huggingface.co — Hosts and serves models for PyTorch, TensorFlow and JAX, and holds the largest public collection of open models and datasets. Free up to 30,000 input characters.

Evaluating and observing models

• Arize AX — Platform for evaluating and observing AI applications and agents. Free tier includes 25,000 spans a month.
• Braintrust — Evaluation, prompt playground and dataset management for generative AI. 1,000 private evaluation rows a week.
• Comet Opik — Open-source platform for evaluating and shipping LLM applications across development and production.
• Future AGI — Open-source platform for evaluating, tracing and improving LLM and agent applications.
• Keywords AI — Monitoring across 200-plus models behind one API format. 10,000 requests free.
• Langfuse — Open-source LLM engineering platform for debugging, analysing and iterating on prompts and traces.
• LangWatch — Platform for measuring and monitoring the reliability of LLM applications.
• Latitude — Open-source (MIT) observability and evaluation for AI applications.
• Maxim — Evaluation and observability with agent simulation and a prompt playground.
• telemetry.dev — Observability for AI applications built on OpenTelemetry, tracing model calls and tool steps.
• Arize AI — Machine learning observability for monitoring models in production and diagnosing drift and data quality problems.
• CometML — MLOps platform for experiment tracking, a model registry and production monitoring.
• Maxim AI — Simulates, evaluates and observes AI agents end to end.
• Weights & Biases — MLOps platform for experiment tracking, dataset versioning and model management. Free for personal use.

Applied AI tools

• Audio Enhancer — Removes noise and echo from audio recordings. Unlimited one-click enhancements, no login, supports MP3, WAV and FLAC.
• Clair — Clinical AI reference suite — literature search, clinical summaries, drug interactions, ICD-10 codes. Free for students.
• ReportGPT — Writing assistant that is free to use if you supply your own model API key.
• Transcript LOL — Transcribes audio and video, then summarises it with a language model.
• Zenable — Checks and corrects output from coding assistants such as Cursor, Windsurf and Copilot against your own standards.
• Composio — Integration layer connecting AI agents and LLMs to over 200 external tools.
• Tavily AI — Search API built for AI agents and research tasks, returning results ready to feed to a model.
• What The Diff — AI code review that summarises what a pull request actually changes. 25,000 tokens a month.

Notebooks and data science environments

• Colaboratory — Hosted Python notebooks with free GPU access. The standard free environment for trying out models.
• Datalore — Python notebooks from JetBrains with 10 GB of storage and 120 hours of runtime a month.
• Deepnote — Data science notebook with real-time collaboration, compatible with Jupyter.
• Hex — Collaborative platform for notebooks, data applications and shared analysis.

Currency

Compiled September 2026. Free tiers change frequently and nothing here has been independently verified — check the provider before relying on any allowance described above. Entries cover services with a standing free tier, not free trials.

See also

Generative AI, Large language model, Open-weight model, Inference, Benchmark, and the other sections of the Free Tier Services Directory.