Explore selected AI models for free on Infron. Requests to these models won’t reduce your credit balance, so you can prototype, learn, and build with the Infron API.
Apodex 1.1 Mini is the 35B-parameter model in the Apodex 1.1 family, designed for agentic work in a smaller, locally deployable form. The family targets multi-step tasks involving reasoning and tool use; completing those tasks requires an external execution environment. Novita serves Mini as a text-input, text-output endpoint with a 262,144-token context and maximum output allowance, and lists reasoning, function calling, and structured outputs. These limits describe Novita's hosted endpoint, not measured performance.
Ling 3.1 Flash is an inclusionAI hybrid reasoning model. Novita describes it as having 560B total parameters with 25B activated, and lists reasoning and function calling for text workloads. Although the catalog's model description mentions a 1M-token context, its structured hosted limits are 262,144 context tokens and 32,768 output tokens. This configuration uses those hosted limits; the broader model-level context claim is not treated as availability on this endpoint.
Atria Dawn Preview is an agentic language model developed by the Shanghai Artificial Intelligence Laboratory for scientific research and engineering. Its official model card describes a GLM-5.2-based mixture-of-experts foundation with 744 billion parameters. It is designed to sustain multi-step work: gathering evidence, planning experiments, writing and running code, inspecting results, and revising failed approaches using environmental feedback. The technical report describes a Verifiable Experience Pipeline that links tool interactions to executable environments and externally checked outcomes. Intended applications include deep research, software and machine-learning development, structured document delivery, and security analysis in authorized environments. These workflows require an external agent harness to execute tools and return observations. The hosted model accepts text input and produces text output, with a 256,000-token context window and a maximum generation limit of 65,536 tokens. It supports Chat Completions, Messages, and Responses interfaces. The release includes MIT-licensed model weights and a separate FP8 checkpoint; this does not establish the precision used by the hosted endpoint. As a preview model, its suitability should be checked against representative tasks; published benchmark results are developer-reported evaluations, not guarantees for individual deployments. Learn more at https://api.atria-asi.ai/docs.
Infron offers free variants of selected models. Free variants carry the :free suffix (for example deepseek/deepseek-v4-flash:free) and are billed at $0.00 per token. Requests to free models never deduct from your credit balance. Learn more: https://infron.ai/docs/overview/free-models
Infron offers free variants of selected models. Free variants carry the :free suffix (for example deepseek/deepseek-v4-flash:free) and are billed at $0.00 per token. Requests to free models never deduct from your credit balance. Learn more: https://infron.ai/docs/overview/free-models
Infron offers free variants of selected models. Free variants carry the :free suffix (for example deepseek/deepseek-v4-flash:free) and are billed at $0.00 per token. Requests to free models never deduct from your credit balance. Learn more: https://infron.ai/docs/overview/free-models
Infron offers free variants of selected models. Free variants carry the :free suffix (for example deepseek/deepseek-v4-flash:free) and are billed at $0.00 per token. Requests to free models never deduct from your credit balance. Learn more: https://infron.ai/docs/overview/free-models