DeepSeek-V4 Pro is the high-performance DeepSeek V4 model, strengthening complex reasoning, coding, analysis, and agentic workflows for production use.

GLM-5 is an advanced open-weight language model designed for developers tackling demanding problems. It excels in long-context reasoning, multi-step tool orchestration, and complex systems engineering, making it a strong fit for advanced agents and applications that need high-level cognition.

GLM-5.1 is Zhipu's latest flagship model, with major improvements in coding and long-horizon tasks, able to plan, execute, and iterate autonomously for extended work.

Gemini-3-Pro-Image-Preview (Nano Banana Pro) is a high-performance image generation and editing model built on Gemini 3 Pro. It offers stronger multimodal understanding and grounded semantic reasoning, generates structured visual content quickly, supports search-based real-time knowledge grounding, and performs especially well on text rendering, multi-image consistency, and identity preservation.

Gemini 3.1 Flash Image Preview (officially marketed as Nano Banana 2) offers professional-grade visual intelligence with lightning-fast efficiency and grounded generation. It is the speed-optimized counterpart to Gemini 3 Pro Image, designed for scale and broad developer use cases.

Gemini 3.1 Pro Preview further improves the Gemini 3 Pro family in performance and reliability, with better reasoning, higher token efficiency, and more grounded, fact-consistent output. It is optimized for software engineering behavior, tool use, and dependable multi-step agent workflows.

GPT-5.4 is our frontier model for complex professional work. Supported reasoning effort options are none (default), low, medium, high, and xhigh.

GPT-Image-2 is OpenAI's next-generation image generation and editing model, improving image quality, instruction following, text rendering, and iterative editing.

Claude Fable 5 is Anthropic’s first publicly available Mythos-class flagship model, built for advanced reasoning, long-horizon agent workflows, software engineering, and scientific research. With ultra-long context, multi-step planning, autonomous task execution, and powerful tool use, it excels in demanding professional applications.

Claude Haiku 4.5 is Anthropic's lightweight high-performance model with fast responses, low latency, and strong cost efficiency for support chat, content generation, extraction, and real-time interaction.

Claude Opus 4.6 is Anthropic's latest frontier reasoning model. It uses adaptive thinking, dynamically deciding when and how much to think. At the default high effort level, it almost always thinks; at lower effort levels, it may skip thinking for simple problems. Note: the minimum cache token threshold for claude-opus-4-6 has increased from 1,024 to 4,096 tokens.

Claude Opus 4.7 is Anthropic's next-generation frontier model, improving complex reasoning, tool use, coding, and agentic task performance for demanding business use cases.

Claude Opus 4.8 is Anthropic’s flagship AI model, delivering exceptional performance in coding, advanced reasoning, research, and agentic workflows. With strong long-context understanding, multi-step problem solving, and tool-use capabilities, it is ideal for software development, data analysis, content creation, and professional applications.

Claude Sonnet 4.6 delivers frontier intelligence at scale for coding, agents, and enterprise workflows. It excels at complex multi-step tasks, iterative software development, end-to-end project execution, financial analysis, cybersecurity workflows, and computer-use scenarios, with improved consistency, autonomy, and production readiness.

Gemini 3 Flash Preview is Google's latest well-balanced model, with strong performance across speed, scale, and frontier intelligence.

Qwen3.5 native vision-language Plus model built on a hybrid architecture that combines linear attention with sparse MoE for higher inference efficiency. Across many benchmarks, the 3.5 series reaches performance comparable to top frontier models, with major gains over the 3 series in both text-only and multimodal tasks. This release is functionally equivalent to the qwen3.5-plus-2026-02-15 snapshot.

GLM-5.2 is Zhipu AI's new flagship open-source model, built for long-horizon tasks and coding. It features a 1M-token lossless context window and can autonomously complete the entire software delivery cycle from development to deployment. It ranks first among globally available models on Code Arena and scores only 1% behind Claude Opus 4.8 on FrontierSWE.

Kimi K2.5 is Kimi's smartest model so far, achieving open-source SoTA results across agent tasks, coding, vision understanding, and broad general intelligence benchmarks. It is also Kimi's most versatile model so far, with a native multimodal architecture that supports vision and text input, thinking and non-thinking modes, and both chat and agent workflows.

MiniMax M2.5 significantly improves performance in high-level productivity scenarios such as coding, search-tool use, and office workflows, reaching or refreshing industry SoTA. It notably improves task decomposition and reasoning efficiency, and completes SWE-Bench Verified tasks 37% faster than the previous M2.1 generation.

MiniMax M2.7 is MiniMax's next-generation general model with stronger coding, tool use, office productivity, and complex task handling.