Apodex: Apodex 1.1 Mini (free)

Apodex 1.1 Mini is a 35-billion-parameter reasoning-first language model from Apodex AI, post-trained from the Qwen3.5-35B-A3B mixture-of-experts base for agentic research, forecasting, and professional workflows. It is designed to work with files, data, code, and tools over long horizons, with native function calling, chain-of-thought style reasoning exposed via compatible parsers, and a 262,144-token context window suited to large documents and multi-step tasks.

The open-weight checkpoint is released on Hugging Face under Apache 2.0 and can be served with SGLang or vLLM using Qwen3.5 chat and tool-call templates. The Apodex 1.1 technical report reports strong agentic results for the mini variant when paired with the Agent Team harness, including 50.2 on FrontierFinance, 51.7 on FrontierScience-Research, and 27.7 on APEX-Agents, with consistent gains over a ReAct-only setup. The same weights are also available as a zero-cost OpenRouter endpoint (apodex/apodex-1.1-mini:free) for OpenAI-compatible chat completions with tools and structured outputs, subject to provider rate limits.

Apodex positions 1.1 Mini as the cost- and latency-optimized sibling to the 397B Apodex 1.1 flagship: tuned for high-volume classification, extraction, drafting, and agent loops while retaining competitive complex-work performance at compact scale. Recommended serving settings for agentic use include temperature 1.0, top_p 0.95, repetition_penalty 1.05, and long max output lengths when streaming is enabled.