Leaderboard
AI model leaderboard
Dynamic ranking of frontier models on how well they handle real editorial tasks — chat, coding, writing, and everyday work — scored by our monthly two-way comparison rounds.
| Rank | Model | Hub Score | Test sessions | API price / 1M | Context |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.5 Anthropic · Proprietary | 1737 +12 / -12 | 2,431 | $3.00 / $15.00 | 200K |
| 2 | GPT-5 OpenAI · Proprietary | 1702 +11 / -11 | 3,102 | $1.25 / $10.00 | 400K |
| 3 | Claude Opus 4.1 Anthropic · Proprietary | 1701 +18 / -18 | 1,466 | $15.00 / $75.00 | 200K |
| 4 | Gemini 2.5 Pro Google · Proprietary | 1681 +13 / -13 | 2,894 | $1.25 / $10.00 | 1M |
| 5 | Grok 4 xAI · Proprietary | 1641 +16 / -16 | 1,782 | $3.00 / $15.00 | 256K |
| 6 | GPT-5 mini OpenAI · Proprietary | 1630 +14 / -14 | 2,238 | $0.25 / $2.00 | 400K |
| 7 | DeepSeek V3.2 DeepSeek · Open source | 1624 +15 / -15 | 2,611 | $0.28 / $0.42 | 128K |
| 8 | Gemini 2.5 Flash Google · Proprietary | 1604 +10 / -10 | 3,567 | $0.30 / $2.50 | 1M |
| 9 | Qwen3-Max Alibaba · Open source | 1601 +17 / -17 | 1,925 | $1.20 / $6.00 | 256K |
| 10 | Kimi K2 Moonshot AI · Open source | 1571 +16 / -16 | 2,054 | $0.60 / $2.50 | 256K |
| 11 | Llama 4 Maverick Meta · Open source | 1533 +14 / -14 | 2,880 | $0.20 / $0.60 | 1M |
| 12 | Mistral Large 3 Mistral AI · Proprietary | 1529 +15 / -15 | 1,733 | $2.00 / $6.00 | 256K |
How this works: Hub Score is our editorial Elo — the same set of real tasks goes to every model each month, our editors compare outputs two at a time, and the score moves as rounds complete. The +a/-b range is the confidence interval. API prices are public list prices per 1M tokens (input/output), approximate — always confirm on the vendor's page. We run this ourselves; it is not a vendor benchmark and we are not affiliated with any lab or arena.