Free AI Model Status
Compare free AI models for coding, chat, JSON output, multimodal tasks, and agent workflows. This live status page tracks OpenRouter free model availability, context length, tool support, JSON support, and a machine-readable probe feed refreshed from Hostinger.
Current Free Model Catalog
| Model | Use | Context | Modalities | Tools | JSON |
|---|---|---|---|---|---|
| baidu/cobuddy:free | Coding and tool-style tasks | 131K | Text | Yes | No |
| openrouter/owl-alpha | Agent/router work | 1M | Text | Yes | Yes |
| nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | Multimodal reasoning | 256K | Text, image, audio, video | Yes | No |
| poolside/laguna-xs.2:free | Coding and agents | 131K | Text | Yes | No |
| poolside/laguna-m.1:free | Coding and agents | 131K | Text | Yes | No |
| inclusionai/ling-2.6-1t:free | General and structured output | 262K | Text | Yes | Yes |
| tencent/hy3-preview:free | General | 262K | Text | Yes | No |
| baidu/qianfan-ocr-fast:free | OCR and image text extraction | 65K | Text, image | No | No |
| google/gemma-4-26b-a4b-it:free | Multimodal general use | 262K | Text, image, video | Yes | Yes |
| google/gemma-4-31b-it:free | Multimodal general use | 262K | Text, image, video | Yes | Yes |
| google/lyria-3-pro-preview | Audio generation preview | 1M | Text, image to audio | No | Yes |
| google/lyria-3-clip-preview | Audio clip preview | 1M | Text, image to audio | No | Yes |
| nvidia/nemotron-3-super-120b-a12b:free | General and reasoning | 262K | Text | Yes | Yes |
| minimax/minimax-m2.5:free | General and agent tasks | 196K | Text | Yes | Yes |
| openrouter/free | Free model router | 200K | Text, image | Yes | Yes |
| liquid/lfm-2.5-1.2b-thinking:free | Small reasoning | 32K | Text | No | No |
| liquid/lfm-2.5-1.2b-instruct:free | Small instruction tasks | 32K | Text | No | No |
| nvidia/nemotron-3-nano-30b-a3b:free | General | 256K | Text | Yes | No |
| nvidia/nemotron-nano-12b-v2-vl:free | Multimodal general use | 128K | Text, image, video | Yes | No |
| qwen/qwen3-next-80b-a3b-instruct:free | General and structured output | 262K | Text | Yes | Yes |
| nvidia/nemotron-nano-9b-v2:free | General and structured output | 128K | Text | Yes | Yes |
| openai/gpt-oss-120b:free | General and reasoning | 131K | Text | Yes | No |
| openai/gpt-oss-20b:free | General | 131K | Text | Yes | No |
| z-ai/glm-4.5-air:free | General | 131K | Text | Yes | No |
| qwen/qwen3-coder:free | Coding and agent tasks | 262K | Text | Yes | No |
| cognitivecomputations/dolphin-mistral-24b-venice-edition:free | General and creative | 32K | Text | No | Yes |
| meta-llama/llama-3.3-70b-instruct:free | General | 65K | Text | Yes | No |
| meta-llama/llama-3.2-3b-instruct:free | Small general tasks | 131K | Text | No | No |
| nousresearch/hermes-3-llama-3.1-405b:free | General | 131K | Text | No | No |
Live Feed Is Running
The Hostinger probe worker refreshes the JSON feed every 15 minutes. It combines the current OpenRouter free-model catalog with lightweight checks for latency, empty responses, rate limits, and JSON-following behavior, so readers can quickly see which free AI models are practical for real work.
Still Being Improved
The public table is a directory, not a full benchmark yet. The next step is broader probes by task type, such as coding, chat, structured JSON, tool use, and agent routing.
How To Use This Status Page
Use this page as a quick triage tool before sending work to a free AI model. For coding and agent workflows, start with models that show tool support and larger context windows. For structured output, prefer models marked with JSON support. For OCR, multimodal, audio, or video workflows, check the modalities column before relying on a model in production.
Free model availability can change without much warning. Providers may rate limit, rotate models, return empty responses during busy periods, or temporarily remove free access. The live feed is designed to catch those changes faster than a static blog post can.
What The Checks Mean
Context is the advertised token window. Tools means the model claims support for function or tool-style calls. JSON means the model advertises or demonstrates structured output behavior. The feed also tracks whether a model is catalog-only, responding normally, rate limited, or returning unusable output.
This is not a paid benchmark lab. It is a practical operations page for developers, small businesses, and AI users who need to know which free models are likely to work right now.
Live JSON Feed
The live probe worker is now publishing a public JSON feed that Hermes and other tools can read:
https://hermes.libertas.mobi/free-model-status/recommendations.json