Benchmarks

Top models across a combined benchmark plus LMArena, Epoch AI, τ-bench, EQ-Bench, the UGI Leaderboard, SpeechMap, and NanoGPT benchmark categories.

Combined

Equal-weight blend of LMArena Overall, Epoch Capabilities Index, NanoGPT Usage Share. Each source is min-max normalized to 0-100 across its current leaderboard and weighted at one third. Missing or unavailable source entries contribute 0.

Top 20 price vs performance

X-axis: $/M blended tokens

1.

72.8%

2.

57.6%

3.

Claude Opus 5.5
Anthropic logo

by Anthropic

56.0%

4.

Claude Fable 5.2
Anthropic logo

by Anthropic

34.8%

6.

Claude Fable 5.1
Anthropic logo

by Anthropic

31.4%

7.

Claude Fable 5
Anthropic logo

by Anthropic

25.6%

8.

25.2%

9.

GPT 6 Astra
OpenAI logo

by OpenAI

22.4%

10.

GPT 6.1 Sol
OpenAI logo

by OpenAI

21.1%

Weighted blend of latest source snapshots

NanoGPT Composite

Includes LMArena data, CC BY 4.0; Epoch AI data, CC BY 4.0. Filtered and ranked to models on NanoGPT.