Penny Owl AI SavingsBack to LeaderboardAI Brain Leaderboard
Which language models are smartest, fastest — and cheapest per million tokens.
- Models ranked
- 65ranked & buyable
- Organisations voting
- 77 boards
- Price quotes tracked
- 203
- Typical price spread
- 1.2×
What this leaderboard ranks
This leaderboard ranks AI language models ("brains") by combined quality, blending matching results from seven organisations — Artificial Analysis, Arena, LiveBench, Epoch AI, Vals AI, Design Arena and Rapidata — with one equal vote each. The blend covers intelligence tests plus coding/design and SVG specialist tests. Beside every rank sit the cheapest verified input and output prices per million tokens, which can come from different sellers.
Prices are per million tokens: input is what you send (prompts and context), output is what the model generates. The two columns can have different cheapest sellers — price a full request with one seller before deciding.
Top of the board
| Rank | Model | Input / 1M tokens · cheapest | Output / 1M tokens · cheapest |
|---|---|---|---|
| 1 | Claude Opus 5.5 | $4.00DeepInfra | $20.00DeepInfra |
| 2 | Claude Fable 5.1 | $10.00Native | $50.00Native |
| 4 | Claude Opus 5 | $5.00DeepInfra | $25.00DeepInfra |
| 5 | Claude Fable 5 | $10.00DeepInfra | $50.00DeepInfra |
| 6 | Claude Sonnet 5.5 | $2.00DeepInfra | $10.00DeepInfra |
| 23 | Claude Opus 4.8 | $4.29Poe | $21.46Poe |
| 29 | Claude Opus 4.7 | $4.29Poe | $21.46Poe |
| 3 | GPT-6 Astra | $10.00Atlas Cloud | $50.00Atlas Cloud |
| 7 | GPT-5.6 Sol | $2.00OpenRouter | $10.00OpenRouter |
| 8 | GPT-6.1 Sol | $2.00Atlas Cloud | $10.00Atlas Cloud |
Cheapest verified offer per column, from the same catalog as the main chart. Click a model for every seller's price and settings.
What makes it special
Brain is the only leaderboard with a Speed column: a separate combined standing from Artificial Analysis 500-token answer estimates and Design Arena frontend timings. Speed never changes the quality rank — it sits beside it, so a slightly-less-famous model that answers twice as fast and costs half as much is easy to spot.
What it is good at
Choosing a brain for real work: balance quality, answer speed and token cost together instead of picking the famous name and paying whatever its home seller charges.
The three leaderboards, side by side
Compare all aggregators →Same rules on every board: one equal vote per organisation, at least two organisations to rank, and price never changes a quality rank. Bar length is relative to the strongest board on that line.
Counts are calculated from the same catalog as the charts, at page render. A wide spread is not a quality score — it is how much the sellers disagree about price.