Frontier · 2026-10-31
Where the capability curve stands
The strongest models, what they cost, and how fast the same capability gets cheaper.
01 · Leading model
97.0
Claude Opus 5.5 · consensus index
02 · What the frontier costs
$4 / $20
No. 1 · input / output per million tokens
03 · Cost of intelligence
713×
GPT-4-level capability since 2023-03
04 · Civilization
K 0.731
on the Kardashev scale
01 — Models
21 benchmarks · 10 organizationsModel leaderboard
- 01Claude Opus 5.597.0
- 02Kimi K3.1 Pro96.9
- 03GPT-6 Astra96.3
- 04Gemini 4 Argon95.2
- 05Claude Fable 5.195.1
Consensus index, not accuracyBrowse by capability →
02 — Prices
What the frontier costs
- Kimi K3.1 Pro
- Gemini 4 Argon
- Claude Opus 5.5
- GPT-6 Astra
- Claude Fable 5.1
InputOutputUSD per million tokens · log scale
Compare two models →
03 — Index · Cost of intelligence
The same intelligence, 713× cheaper.
In 2023-03, GPT-4-level capability cost $37.50 per million tokens. Today Qwen3.7 Flash does as well for $0.053.
Today · per million tokens
- GPQA ≥ 35%
- $0.053
- 713×
- GPQA ≥ 50%
- $0.053
- 114×
- GPQA ≥ 75%
- $0.053
- 499×
- GPQA ≥ 90%
- $0.20
- 23×
157 models · as of 2026-11-01See how the price falls →
04 — Civilization
K = 0.731
Humanity's primary power on the Kardashev scale, where a Type I civilization is 1.
Leaderboard scores are a consensus index, not accuracy. See the leader's benchmark evidence