Changelog

Add Claude Opus 4.6

Add GPT-5.3-Codex

From now on, at /models page and each model detail page, users are allowed to compare the costs and model functionality details at once within 3 models at most.

Enjoy.

Introducing Apertis Coding Plan

Today, we're excited to announce the launch of Apertis Coding Plan — a new subscription-based pricing model designed specifically for AI coding tools and agent-based workflows.

You can learn more about the available plans here: 👉 https://apertis.ai/subscribe

---

Why we built Coding Plan

Over the past months, we've seen a clear pattern among developers using AI coding tools and agents.

Tools like Claude Code, OpenClaw, Cline, and OpenCode dramatically increase productivity — but they also introduce a new problem: unpredictable token costs.

Agent-based workflows send far more than just a short prompt. Each request often includes full file context, conversation history, tool outputs, and system instructions. What looks like a derived request can easily become a very large API call behind the scenes.

In a pay-per-token world, this makes costs difficult to predict and hard to control — especially for high-frequency, large-context workloads.

Coding Plan was created to address this exact pain point.

---

A pricing model designed for agents and coding tools

Instead of charging purely by token usage, Coding Plan introduces a fixed monthly subscription with a rate-based quota system.

Each model on Apertis has a defined rate multiplier. Every request consumes quota based on the model used — lightweight models consume less, while premium models consume more. This allows developers to consciously allocate usage between everyday tasks and heavier reasoning workloads.

Importantly, this system is not credit-based and not a fixed call limit. It's designed to give developers flexibility while keeping costs predictable.

---

Free cache hits for repeated context

Agent workflows frequently re-read the same context: source files, repository structure, or intermediate tool outputs.

With Coding Plan, cache hits are completely free. Repeated prompts and context reads do not consume quota and do not incur token charges. This significantly reduces costs for iterative coding and agent polling patterns, where cache usage can represent a large portion of total traffic.

---

Free unlimited models for daily tasks

Coding Plan also includes a set of models that can be used without consuming any quota:

• Gemini 3 Flash
• Gemini 2.5 Flash
• GPT OSS 120B
• DeepSeek V3.1

and more in the future ✨ These models are well-suited for tasks like autocomplete, formatting, lightweight refactors, and quick lookups — allowing developers to reserve quota for more complex reasoning when needed.

---

Staying online when quota runs out

For workloads that occasionally exceed monthly expectations, Coding Plan offers an optional On-Demand Usage fallback.

When enabled, requests continue at approximately 50% of standard market token prices, with a configurable monthly spending cap. Usage automatically stops once the cap is reached, preventing unexpected overage charges.

This ensures workflows don't abruptly stop while still maintaining cost control.

---

Built to work with existing tools

Coding Plan is compatible with major AI coding tools and agent frameworks, including OpenClaw, Claude Code, Cline, OpenCode, Aider, Cursor, and more.

For most users, migration is straightforward:

• Update the API base URL
• Replace the API key

No changes to request formats, SDKs, or workflows are required.

---

A single API key, multiple providers

With Coding Plan, developers gain access to 30+ models through a single API key, spanning multiple providers and model families:

Claude, GPT, Gemini, DeepSeek, Kimi, GLM, MiniMax, and others.

Apertis also supports automatic failover between providers, ensuring requests continue even if a specific upstream service becomes unavailable.

---

Getting started

Apertis Coding Plan is available starting today.

For full plan details, pricing, usage estimates, and migration guidance, visit: 👉 https://apertis.ai/subscribe

As always, we welcome feedback and questions from the community as we continue improving the platform.

Add Qwen3 Coder Next