Thị trường mô hình

DeepSeek V4.1 Flash is a 552B-parameter MoE model featuring a novel Causal-Encoder-Decoder architecture with asymmetric inputs and outputs. With an input activation of only 8B and an output activation of 16B, its cost is significantly lower than existing models of the same size.

Google's flagship reasoning model for complex software engineering, deep math and science, and long-horizon autonomous agents

Fast Claude model for responsive assistance, classification, and lightweight agents

Gemini Nano Banana 2.1 is Google's efficient image generation and conversational intelligent photo editing model.

GPT-6.1 Sol delivers near-Astra performance at a lower cost for complex coding, computer use, and professional work. Compare it with Astra on your tasks to assess the tradeoff between quality and cost.

Claude Sonnet 5.5 offers the best combination of speed and intelligence, but it introduces some breaking changes in parameter passing.

Claude Opus 5.5 is built for long-running agentic coding and knowledge work

GPT-6 Luna is OpenAI's most responsive and efficient model designed for high-intensity tasks.

GPT-6 Sol is built for complex coding and agentic workflows.

Grok 4.7 is x.ai's next-generation flagship model: agentic tool calling, minimal hallucinations, and configurable reasoning.

A next-generation omni-modal large model with native upport for unified understanding across text, vision and audio.

A next-generation large model advancing toward production-grade intelligence, featuring comprehensive upgrades in coding, agent, and multimodal capabilities.

GPT-6 Astra excels at executing multi-step workflows across code, browsers, and professional software. In multiple evaluations, Astra achieved stronger results while using significantly fewer output tokens—and despite its higher per-token pricing, its estimated API cost per task was lower than that of earlier models.

Gemini 3.8 Flash is engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows—all with the speed and cost efficiency of Flash.

Claude Fable 5.1 costs just one-quarter as much to read from the cache as Claude Fable 5, while delivering stronger capabilities for long-running agentic coding, multi-step research, and working with documents, spreadsheets, and presentations.

GLM-5.3-Flash is a native multimodal model, with certain capabilities on par with Claude Opus 4.8.

Qwen3.8-Flash combines powerful understanding and generation capabilities with exceptional response speed, enabling it to process ultra-long documents, code repositories, and complex conversations in a single pass.

GLM-5.3 improves the coding experience by 50%, with select cybersecurity capabilities on par with Mythos 5.

The most notable change in V4 is that a million-token ultra-long context window has become standard across all official services, along with deep optimization for various agent tools.

The most notable change in V4 is that million-word-long context has become standard across all official services, with deep optimization for various agent tools.

Gemini 3.7 Flash is the next iteration in the Gemini 3 series of highly-capable, natively multimodal, reasoning models.

Grok 4.6 builds on Grok 4.5 with a particular focus on long-running agents and more ambitious interactive and visual work. It stays with complex tasks across many steps, whether researching a topic, analyzing information, working across a codebase, or turning an idea into a polished application or work artifact.

The 24-trillion-parameter MoE flagship model demonstrates a comprehensive leap in programming and office capabilities, capable of independently completing full projects through autonomous programming within a matter of days.

October 17 update of DeepSeek V4.1 Flash with improved reasoning, coding, agentic tasks, and visual understanding

October 17 update of DeepSeek V4.1 Flash with improved reasoning, coding, agentic tasks, and visual understanding

Opus 5 is designed to be used every day: it works more efficiently than other models. It’s the new default model on Claude Max, and the strongest model on Claude Pro.[Check out the official intro](https://www.anthropic.com/news/claude-opus-5)

Gemini 3.5 Flash-Lite is a low-latency, cost-effective multimodal model specifically optimized for high-throughput, low-cost sub-agent tasks and document parsing.

Gemini 3.6 Flash is faster and more cost-effective. It's designed for the era of intelligent agents, excelling in code generation, agent execution, and spatial reasoning. This model is particularly well-suited for agent tasks that require frequent and complex programming loops and iterations, delivering outstanding efficiency.

Kimi K3 has 2.8 trillion parameters. It is built on the Kimi hybrid linear attention mechanism (Kimi Delta Attention) and Attention Residuals technology. With native support for visual understanding and a context window of up to 1 million tokens, it is designed for cutting-edge intelligent scenarios such as software engineering, knowledge work, and deep reasoning.