PolicyEngine

Models

Each organization appears at its frontier tier as of its elicitation date, and superseded models stay in the panel — so generation-to-generation shifts are directly observable. Tightness is the model's average 90 percent interval-width rank across the 13 canonical quantities (1 = narrowest of 31).

GPT-5.4
Organization
OpenAI
Wave
April 2026
Tightness
#15.3
GPT-5.4 mini
Organization
OpenAI
Wave
April 2026
Tightness
#23.1
GPT-5.4 nano
Organization
OpenAI
Wave
April 2026
Tightness
#18.8
Claude Opus 4.7
Organization
Anthropic
Wave
April 2026
Tightness
#9.5
Claude Sonnet 4.6
Organization
Anthropic
Wave
April 2026
Tightness
#13.7
Claude Haiku 4.5
Organization
Anthropic
Wave
April 2026
Tightness
#9.3
Gemini 3.1 Pro
Organization
Google
Wave
April 2026
Tightness
#11.2
Gemini 3 Flash
Organization
Google
Wave
April 2026
Tightness
#12.9
Gemini 3.1 Flash-Lite
Organization
Google
Wave
April 2026
Tightness
#13.9
Grok 4.20
Organization
xAI
Wave
April 2026
Tightness
#23.8
Grok 4.1 Fast
Organization
xAI
Wave
April 2026
Tightness
#19.2
GPT-5.5
Organization
OpenAI
Wave
July 2026 frontier
Tightness
#15.1
Claude Fable 5
Organization
Anthropic
Wave
July 2026 frontier
Tightness
#6.8
Claude Opus 4.8
Organization
Anthropic
Wave
July 2026 frontier
Tightness
#11.0
Claude Sonnet 5
Organization
Anthropic
Wave
July 2026 frontier
Tightness
#9.8
Gemini 3.5 Flash
Organization
Google
Wave
July 2026 frontier
Tightness
#13.3
Grok 4.3
Organization
xAI
Wave
July 2026 frontier
Tightness
#18.7
DeepSeek V4 Pro
Organization
DeepSeek
Wave
July 2026 independent labs
Tightness
#20.4
Qwen 3.7 Max
Organization
Alibaba
Wave
July 2026 independent labs
Tightness
#20.3
Kimi K2.6
Organization
Moonshot AI
Wave
July 2026 independent labs
Tightness
#22.7
GLM-5.2
Organization
Zhipu AI
Wave
July 2026 independent labs
Tightness
#20.8
MiniMax M3
Organization
MiniMax
Wave
July 2026 independent labs
Tightness
#18.8
GPT-5.6 Sol
Organization
OpenAI
Wave
July 2026 GPT-5.6
Tightness
#17.4
GPT-5.6 Luna
Organization
OpenAI
Wave
July 2026 GPT-5.6
Tightness
#26.2
GPT-5.6 Terra
Organization
OpenAI
Wave
July 2026 GPT-5.6
Tightness
#13.8
Grok 4.5
Organization
xAI
Wave
July 2026 late
Tightness
#18.8
Kimi K3
Organization
Moonshot AI
Wave
July 2026 late
Tightness
#18.8
Gemini 3.6 Flash
Organization
Google
Wave
July 2026 late
Tightness
#7.5
Claude Opus 5
Organization
Anthropic
Wave
July 2026 late
Tightness
#13.0
Qwen 3.8 Max
Organization
Alibaba
Wave
August 2026
Tightness
#19.4
Inkling
Organization
Thinking Machines
Wave
August 2026
Tightness
#12.7

12,090 successful runs · elicited April through August 2026 · v4 prompts · 15 runs per model-quantity cell. Code · Raw responses · Paper (PDF)

Code and dataElicited April through August 2026 · 31 models · v4 prompts