Model List

Claude Sonnet 5 is the next generation of Anthropic's Sonnet model family. It is a drop-in upgrade for Claude Sonnet 4.6 with three behavior changes: adaptive thinking is on by default, manual extended thinking now returns a 400 error (it was deprecated on Claude Sonnet 4.6)

Nano Banana Lite is Gemini's most lightweight image generation model, capable of quickly generating 1K resolution images (imageSize=1K). It is especially suitable for scenarios where image generation speed is a priority. Image price: $0.0336 per piece.

Nano Banana Lite is Gemini's most lightweight image generation model, capable of quickly generating 1K resolution images (imageSize=1K). It is especially suitable for scenarios where image generation speed is a priority. Raw image price: $0.0336 per image

GLM-5.2 is the flagship model for the era of long tasks. It supports a truly usable 1M context, can handle project-level engineering context, executes long-range tasks more stably, follows engineering standards more reliably, and can complete the entire development process from requirements to multi-end deployment in one go.

Kimi K2.7 Code follows instructions more reliably in long contexts and achieves a higher success rate in completing programming tasks. It supports input in text, images, and videos, and only supports thinking mode, as well as dialogue and agent tasks.

Kimi K2.7 Code HighSpeed is the high-speed version of Kimi K2.7 Code. It uses the same model as Kimi K2.7 Code but delivers an output speed of approximately 180 tokens per second, and can reach up to 260 tokens per second in short-context scenarios, offering an enhanced programming experience.

Anthropic's most capable widely released model, for the most demanding reasoning and long-horizon agentic work，Compared to the previous generation Opus4.8, the execution efficiency has been significantly improved.

Claude Opus 4.8 offers improved long-term context handling, reduced synthesis, and better compression recovery. Inference calibration ensures more reliable performance across all levels of effort in various domains. Tool activation is less frequent, resulting in fewer tool calls needed to skip tasks.

The new dynamic thinking model of Claude-Opus-4-8 automatically determines whether deep thinking is required based on prompts. Suffixes such as -max, -high, -medium, and -low indicate the level of reasoning.

The new dynamic thinking model of Claude-Opus-4-8 automatically determines whether deep thinking is required based on prompts. Suffixes such as -max, -high, -medium, and -low indicate the level of reasoning.

The new dynamic thinking model of Claude-Opus-4-8 automatically determines whether deep thinking is required based on prompts. Suffixes such as -max, -high, -medium, and -low indicate the level of reasoning.

The new dynamic thinking model of Claude-Opus-4-8 automatically determines whether deep thinking is required based on prompts. Suffixes such as -max, -high, -medium, and -low indicate the level of reasoning.

The new dynamic thinking model of Claude-Opus-4-8 automatically determines whether deep thinking is required based on prompts. Suffixes such as -max, -high, -medium, and -low indicate the level of reasoning.

The new dynamic thinking model of Claude-Opus-4-8 automatically determines whether deep thinking is required based on prompts. Suffixes such as -max, -high, -medium, and -low indicate the level of reasoning.

Gemini 3.1 Flash Image is a more efficient image generation version, optimized for speed and high-usage developer scenarios, with significantly reduced pricing. It also supports a 0.5k (512) resolution and adds new image aspect ratios: 1:4, 4:1, 1:8, and 8:1.

Gemini 3.1 Flash Image is a more efficient image generation version, optimized for speed and high-usage developer scenarios, with significantly reduced pricing. It also supports a 0.5k (512) resolution and adds new image aspect ratios: 1:4, 4:1, 1:8, and 8:1.

Gemini 3.1 Flash Image is a more efficient image generation version, optimized for speed and high-usage developer scenarios, with significantly reduced pricing. It also supports a 0.5k (512) resolution and adds new image aspect ratios: 1:4, 4:1, 1:8, and 8:1.

Gemini 3.1 Flash Image is a more efficient image generation version, optimized for speed and high-usage developer scenarios, with significantly reduced pricing. It also supports a 0.5k (512) resolution and adds new image aspect ratios: 1:4, 4:1, 1:8, and 8:1.

nano-banana-pro (gemini-3-pro-image) is optimized based on the previous generation gemini-2.5-flash-image, offering more professional drawing capabilities. It supports image creation and editing in 1k, 2k, and 4k quality, making it the preferred choice among current drawing models.

nano-banana-pro (gemini-3-pro-image) enables 2K quality drawing.（1~2k=$0.134、4k=$0.24）

nano-banana-pro (gemini-3-pro-image) enables 4K quality drawing.

The high cost-performance Plus model in the Qwen3.7 series has comprehensively upgraded its vision-language capabilities on top of its strong text abilities, while maintaining full agent capabilities in coding, tool usage, and productivity workflows.

Gemini 3.5 Flash can continuously deliver cutting-edge intelligence and is optimized for faster, more cost-effective handling of real-world tasks. Designed for the era of intelligent agents, it excels at large-scale deployment of sub-agents, managing multi-step workflows, and handling long-duration tasks. This model is particularly well-suited for rapid agent cycles that involve complex coding iterations and iterative processes.

Qwen3.7 is a next-generation flagship model designed for the era of intelligent agents, with its core strengths lying in the breadth and depth of its agent capabilities: it excels at a wide range of tasks in programming, office work and productivity, as well as long-term autonomous execution.

gemini-3.5-flash-search is an official web_search online search model with limited concurrency.

Gemini 3.5 Flash enables the forced deep thinking version.

Gemini 3.5 Flash disables deep thinking capability.

Gemini 3.1 Flash-Lite is best suited for handling massive proxy tasks, simple data extraction tasks, and extremely low-latency applications where budget and speed are the primary limiting factors.

GPT-5.5 is our OpenAI frontier model for the most complex professional work. Reasoning.effort supports: none, low, medium (default), high and xhigh.

GPT-5.5 is our OpenAI frontier model for the most complex professional work. Reasoning.effort supports: none, low, medium (default), high and xhigh.