⌕Is ___ SOTA?
isitsota / categories / Reasoning
Reasoning / knowledge · Oct 13, 2026
Who is SOTA in Reasoning?
Gemini 4 Argon & O3 Mini & DeepSeek R1 & Kimi K2 5 & GPT-5 Mini & Llama 3.3 70B Instruct & Gemma 3 27B IT & Phi 4

Hard reasoning and expert-question performance. Saturated exams drop out of the crown automatically. As of Oct 13, 2026, Gemini 4 Argon & O3 Mini & DeepSeek R1 & Kimi K2 5 & GPT-5 Mini & Llama 3.3 70B Instruct & Gemma 3 27B IT & Phi 4 holds Reasoning. On…

OverallOpen weights≤ 32BBest per $
Standings · pareto

The board

History →
#ModelStateConfidenceWhy
1GGemini 4 Argon
Google DeepMind		ParetomediumOn the cost-quality frontier for Reasoning: nothing scores better for less than $4.00 per 1M blended tokens.
1OGPT-6 Astra
OpenAI		Not SOTA · #1mediumNot on the Reasoning cost-quality frontier.
3OO3 Mini
OpenAI		ParetomediumOn the cost-quality frontier for Reasoning: nothing scores better for less than $1.93 per 1M blended tokens.
4GGemini 3.1 Pro Preview
Google DeepMind		Not SOTA · #4mediumNot on the Reasoning cost-quality frontier.
6OGPT-5.4 Pro
OpenAI		Not SOTA · #6mediumNot on the Reasoning cost-quality frontier.
7AClaude Opus 5.5
Anthropic		Not SOTA · #7mediumNot on the Reasoning cost-quality frontier.
7DSDeepSeek R1
DeepSeek		ParetomediumOn the cost-quality frontier for Reasoning: nothing scores better for less than $1.15 per 1M blended tokens.
11OGPT-5.4
OpenAI		Not SOTA · #11mediumNot on the Reasoning cost-quality frontier.
14AClaude Opus 4.6
Anthropic		Near-SOTAmediumNot on the Reasoning cost-quality frontier.
16OGPT-5 Pro
OpenAI		Not SOTA · #16mediumNot on the Reasoning cost-quality frontier.
18OGPT-5.2
OpenAI		Not SOTA · #18mediumNot on the Reasoning cost-quality frontier.
20OGPT-5
OpenAI		Not SOTA · #20mediumNot on the Reasoning cost-quality frontier.
22AClaude Fable 5.1
Anthropic		Near-SOTAmediumNot on the Reasoning cost-quality frontier.
23AClaude Opus 4.5
Anthropic		Not SOTA · #23mediumNot on the Reasoning cost-quality frontier.
25KKimi K2 5
Moonshot AI		ParetomediumOn the cost-quality frontier for Reasoning: nothing scores better for less than $1.00 per 1M blended tokens.
Pareto

Quality against blended price

Gemini 4 ArgonGPT-6 AstraO3 MiniGemini 3.1 Pro PreviewGPT-5.4 ProClaude Opus 5.5DeepSeek R1GPT-5.4Claude Opus 4.6GPT-5 ProGPT-5.2GPT-5
Evidence recipe

What can move this crown