LIVE

Disrupting a Criminal Scam Operation04/08/26 · OpenAI|Anionex/codex-deepseek-vision: 让纯文本模型在 Codex 中无障碍调用内置看图工具（view_image）的方案，附为纯文本 LLM 设计的视觉工具包｜ Let text-only models call Codex's built-in view_image seamlessly, plus a vision toolkit designed for text-only LLMs.01/08/26 · DeepSeek|Ten advances in mathematics and theoretical computer science01/08/26 · OpenAI|Advancing responsible AI across Europe31/07/26 · OpenAI|Building abundant intelligence31/07/26 · OpenAI|DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis31/07/26 · DeepSeek|Google fixed more Chrome bugs in June than over the past two years, thanks to AI31/07/26 · Google|Univé builds an AI-ready workforce31/07/26 · OpenAI|Show HN: What should the GUI for AI agents look like?31/07/26 · OpenAI|Show HN: Distilling DeepSeek into GPT-OSS doesn't transfer censorship. Try it30/07/26 · OpenAI|AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis30/07/26 · OpenAI|OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models30/07/26 · Microsoft|Disrupting a Criminal Scam Operation04/08/26 · OpenAI|Anionex/codex-deepseek-vision: 让纯文本模型在 Codex 中无障碍调用内置看图工具（view_image）的方案，附为纯文本 LLM 设计的视觉工具包｜ Let text-only models call Codex's built-in view_image seamlessly, plus a vision toolkit designed for text-only LLMs.01/08/26 · DeepSeek|Ten advances in mathematics and theoretical computer science01/08/26 · OpenAI|Advancing responsible AI across Europe31/07/26 · OpenAI|Building abundant intelligence31/07/26 · OpenAI|DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis31/07/26 · DeepSeek|Google fixed more Chrome bugs in June than over the past two years, thanks to AI31/07/26 · Google|Univé builds an AI-ready workforce31/07/26 · OpenAI|Show HN: What should the GUI for AI agents look like?31/07/26 · OpenAI|Show HN: Distilling DeepSeek into GPT-OSS doesn't transfer censorship. Try it30/07/26 · OpenAI|AskChem: Claim-Centered Infrastructure for Chemistry Literature Synthesis30/07/26 · OpenAI|OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models30/07/26 · Microsoft|

🏆 The AI Podium

The reference ranking of the best AI models, updated every month.

July 2026 edition

🏆 The AI Podium

Official ranking · July 2026

View methodology See full podium

Generalists

Best all-rounders

Moonshot AI·88/100·$0.95/$4mit

Claude Opus 4.8

Anthropic·85/100·$5/$25

Google·85/100·$2/$12

OpenAI·83/100·$5/$30

DeepSeek·83/100·$0.28/$0.42mit

Code Specialists

SWE-bench + HumanEval

Claude Opus 4.8

Anthropic·87/100·$5/$25

OpenAI·87/100·$5/$30

OpenAI·85/100·$1.25/$10

Moonshot AI·85/100·$0.95/$4mit

Claude Opus 4.7-2

Anthropic·83/100·$5/$25

Vision Specialists

Multimodal & image

Gemini 3.1 Pro+2

Google·81/100·$2/$12

OpenAI·79/100·$5/$30

Claude Opus 4.8

Anthropic·79/100·$5/$25

Google·78/100·$1.25/$5

Alibaba·77/100·$0.78/$3.9

Multilingual Specialists

FR · EN · CN & more

Moonshot AI·88/100·$0.95/$4mit

Alibaba·88/100·$0.78/$3.9

Mistral Large 3

Mistral AI·86/100·$2/$6commercial

Zhipu AI·86/100·$0.5/$2apache-2

DeepSeek·85/100·$0.28/$0.42mit

Open Source

Self-hostable, free license

Mistral Small 3

Mistral AI·78/100·$0.2/$0.6apache-2

Moonshot AI·77/100·$0.95/$4mit

Zhipu AI·68/100·$0.5/$2apache-2

Mistral AI·67/100·$0.6/$0.6apache-2

DeepSeek·66/100·$0.55/$2.19mit

Local · Laptop

Light · < 16 GB RAM

Alibaba·91/100apache-2.0

Alibaba·88/100apache-2.0

Google·88/100apache-2.0

Google·86/100apache-2.0

Alibaba·86/100apache-2.0

Local · Desktop

~24 GB RAM

Gemma 4 26B-A4B

Google·92/100apache-2.0

Google·91/100apache-2.0

Qwen3.6 35B-A3B

Alibaba·90/100apache-2.0

Alibaba·89/100apache-2.0

OpenAI·87/100apache-2.0

Local · Workstation/Server

Big VRAM

DeepSeek V4-Flash

DeepSeek·86/100mit

DeepSeek·85/100mit

OpenAI·85/100apache-2.0

Zhipu AI·77/100mit

Qwen3.5 397B-A17B

Alibaba·76/100apache-2.0

🔬 How is this ranking made?

A transparent algorithm combines official benchmarks, pricing, context window and freshness to score each model.

View full methodology →

📅 Monthly archives

July 2026 (current)