Benchmark rank
AI Metrics · weekly model watch
Which models are leading, and what do they cost?
A small public dashboard for the model signals I care about: independent benchmark scores, Arena preference ratings, and the slow grind of token prices over time.
Freshness
Preview
Models shown
5
Snapshot
2026-07-06
Updating
Weekly
Preview data · connect D1 and source keys for live weekly snapshots
Token economics
Price per 1M tokens over time
Rough heuristic
Benchmark score per blended dollar
This is intentionally blunt: benchmark score divided by blended price per 1M tokens. It is useful for curiosity, not procurement.
Lookup table
Top 10 details
| Rank | Model | Provider | AA score | Arena | Input | Output |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 Thinking | Anthropic | 74.8 | 1,504 | $5 | $25 |
| 2 | GPT-5.5 High | OpenAI | 73.9 | 1,487 | $5 | $20 |
| 3 | Gemini 3.1 Pro | 72.6 | 1,441 | $2.5 | $15 | |
| 4 | DeepSeek V3 | DeepSeek | 69.4 | 1,398 | $0.27 | $1.1 |
| 5 | Qwen 3.7 Plus | Alibaba | 68.8 | 1,439 | $0.4 | $1.6 |
Source health
fallback
healthy
7/6/2026, 12:00:00 AM