Apodex 1.1 Mini Arrives on Novita AI With 256K Context and $0 Launch Pricing

What Apodex 1.1 Mini is

Apodex 1.1 Mini belongs to the Apodex 1.1 series. According to Novita, it uses a Qwen3.5-MoE architecture and runs on dedicated H200 GPU fleets with TP4 tensor parallelism, prefix-cache-aware routing and MTP speculative decoding. Novita notes that these deployment details help developers think about latency and throughput, but they do not replace testing with their own prompts and tool workflows.

The model is text in, text out. Its listed context window is 262,144 tokens (256K), and the listed maximum output is also 262,144 tokens. The catalog lists function calling, structured outputs and reasoning as features. It is served through Novita's serverless API and supports the Chat Completions, Anthropic and Responses endpoints.

How developers can call it

The model ID is apodex/apodex-1.1-mini, and requests go to the base URL https://api.novita.ai/v3/openai. Because the API is OpenAI-compatible, the standard OpenAI Python client only needs a Novita base URL, a Novita API key and the model ID. Novita advises keeping the key in an environment variable rather than in source control. The simplest route is the Chat Completions endpoint, and streaming works by setting stream=True without changing the model ID or authentication.

For tool use, Novita says the catalog lists native function calling. A tool definition tells the model what an application can do, but the application still owns execution, validation, permissions and the final tool result. Novita recommends treating tool arguments as untrusted model output and warns against letting a model-generated argument directly become a shell command, SQL statement or privileged API call.

On reasoning, Novita says controls and response behavior can vary by client and endpoint. It suggests starting with default behavior and checking the current API documentation before relying on a provider-specific reasoning parameter.

Pricing and caveats

As of October 1, 2026, Novita's model catalog and pricing page show both input and output at $0 per million tokens. Novita describes this as a launch promotion, not a guarantee of permanent free access, and says pricing, rate limits and availability can change. It advises checking the live pricing page and the Apodex model page before estimating production costs.

Even at a $0 listed price, Novita recommends setting application-level budgets and rate limits, since large-context requests can use substantial compute and may be subject to account, endpoint or promotional limits. It also suggests recording token usage and keeping a fallback model configured if an application depends on continuous access.

Who it suits

Novita names several candidate uses: coding assistants that inspect long files or repository context, agent workflows that call tools and return structured data, prototypes that benefit from 256K context without self-hosting a large model, and evaluations of MoE inference and long-context behavior on a hosted API.

The model may be a poor fit for teams that need image or audio input, a fixed long-term price contract, or a feature Novita does not list for this endpoint. Novita says a large context does not mean every request should carry the whole workspace, and recommends trimming irrelevant files and bounding tool results. For production evaluation, it suggests comparing answer quality on real prompts, tool-call validity, time to first token, total latency and context trimming behavior.