AI Model Ranking by Functional Category

Because the AI model landscape spans many types (chat, coding, image generation, video generation, voice, open-source...), the npmrundeploy.com team compiles a weekly breakdown by functional category so you can pick the right model for the right job — curated from public arenas/leaderboards and this week's news, not our own automated benchmark.

Week of 2026-07-20

💬 Chat & Writing (General-purpose LLM)

1Claude (Opus / Sonnet)

Maker: Anthropic

Leads several composite leaderboards for answer quality and reliability.

2GPT-5 family

Maker: OpenAI

The largest ecosystem, deeply integrated into ChatGPT and many third-party products.

3Gemini 3 Pro

Maker: Google

Strong at multimodal understanding and long context, integrated with Google Search/Workspace.

4Grok 4

Maker: xAI

Deeply integrated into X (Twitter), strong at real-time information.

💻 Coding & coding agents

1Kimi K3

Maker: Moonshot AI

Topped a major coding arena this week; trial demand spiked sharply.

2GLM-5.1

Maker: Zhipu / Z.AI

An open model with coding performance directly competitive with closed models.

3Claude Code / Claude Opus

Maker: Anthropic

Strongest at agentic coding — understanding and editing large codebases across multiple steps.

4Qwen 3.7

Maker: Alibaba

A popular open model for coding, well balanced between speed and quality.

🧩 Reasoning & AI Agents

1DeepSeek V4

Maker: DeepSeek

Just shipped its stable release this week — strong reasoning, low cost, pressuring industry pricing.

2GLM-5

Maker: Zhipu / Z.AI

Among the top open models for agentic coding on SWE-bench.

3GPT-5 (reasoning mode)

Maker: OpenAI

Strong at long reasoning chains, widely used for research and complex agent tasks.

🎨 Image Generation

1GPT Image 2

Maker: OpenAI

Currently tops the arena leaderboard for generated image quality.

2GPT Image 1.5

Maker: OpenAI

The previous version, still widely used due to lower cost.

3Gemini 3.1 Flash Image

Maker: Google

Fast generation, built into the Gemini/Google ecosystem.

4Midjourney

Maker: Midjourney Inc.

Aesthetic quality still rated highest by designers and artists.

🎬 Video Generation

1Kling v3

Maker: Kuaishou

Currently leads the text-to-video arena leaderboard by score.

2Seedance 2.0

Maker: ByteDance

Close behind the leader, strong at smooth motion and longer durations.

3LTX-2.3 Pro

Maker: Lightricks

Leads the open-weight group for image-to-video generation with audio.

🎙️ Voice (TTS / Speech-to-Text)

1ElevenLabs

Maker: ElevenLabs

Leads in natural voice synthesis and voice cloning, widely used in news-reader/audiobook apps.

2OpenAI Realtime Voice

Maker: OpenAI

Low-latency real-time voice, used for interactive voice assistants.

3Whisper (kế thừa)

Maker: OpenAI

Still a popular open-source choice for speech-to-text.

🧬 Open-source / On-device

1Qwen 3.7

Maker: Alibaba

A well-rounded open model, strong at both coding and multilingual tasks.

2GLM-5.1

Maker: Zhipu / Z.AI

Coding performance competitive with closed models, open license for enterprise use.

3DeepSeek V4 Pro

Maker: DeepSeek

The strongest open model for enterprise agentic coding per SWE-bench.

4Gemma 4

Maker: Google

The only open model family with variants specifically optimized for smartphones.

5Llama 4 Scout

Maker: Meta

Extremely long context window (up to 10M tokens), the most widely deployed in enterprise.