The ranking of AI models in Spanish

Name: Ranking IA — evaluation of frontier models in Spanish
Creator: Sozpic
License: https://creativecommons.org/licenses/by/4.0/

Last updated: September 1, 2026

9 frontier models, 25 real-world tests, 6 families. Pre-registered methodology, automated judges with recusal by lab, and every raw response published. Every cell in the table opens and can be read.

Models

Tests

Runs

Responses

900

1.0

Tiers

Sorted by overall score

ranking.sozpic.com

≥ 90

Claude Fable 5.1

Anthropic

97.3%

Claude Opus 5

Anthropic

97.1%

Claude Fable 5

Anthropic

94.3%

80–89

Kimi K3

Moonshot

89.2%

GPT-5.6 Sol

OpenAI

87.2%

70–79

Grok 4.6

xAI

72.1%

DeepSeek V4 Flash

DeepSeek

72.1%

Gemini 3.1 Pro

Google

71.2%

60–69

DeepSeek V4 Pro

DeepSeek

67.3%

< 60

S ≥ 90 · A 80–89 · B 70–79 · C 60–69 · D < 60

2.0

Table

No empty cells. Every one of them opens.

Overall score: the unweighted mean of the 6 families.

Ranking IA — overall and per-family scores of 9 frontier models in Spanish. Run 2026-09-01_7e3c679, data as of September 1, 2026. Scores as a percentage of the rubric maximum.
#	Model	Lab	Overall	Tier	95% CI	Output price $/1M	Instructions	Truthfulness	Reasoning	Character	Expression	Spanish	Consistency
1	Claude Fable 5.1 claude-fable-5-1	Anthropic	97.3%	S	95.7–98.8	50	98.3%	100%	91.7%	100%	98.4%	95.3%	±2.1
1	Claude Opus 5 claude-opus-5	Anthropic	97.1%	S	95.4–98.7	25	94.4%	99.2%	100%	99.4%	98.1%	91.7%	±3.3
3	Claude Fable 5 claude-fable-5	Anthropic	94.3%	S	92.5–95.9	50	93.1%	95%	95.8%	100%	82.9%	99%	±3.4
4	Kimi K3 accounts/fireworks/models/kimi-k3	Moonshot	89.2%	A	86.5–91.7	15	84.9%	97.5%	77.1%	96.4%	94.8%	84.4%	±6.2
4	GPT-5.6 Sol gpt-5.6-sol	OpenAI	87.2%	A	85.2–89.2	20	100%	93.3%	83.3%	75.6%	86.6%	84.4%	±6
6	Grok 4.6 grok-4.6	xAI	72.1%	B	70.2–73.9	6	65.3%	85.8%	83.3%	60.7%	64.4%	72.9%	±7
6	DeepSeek V4 Flash deepseek-v4-flash	DeepSeek	72.1%	B	69.1–75.4	1.3	71.4%	80%	71.9%	30.4%	96.2%	82.8%	±13
6	Gemini 3.1 Pro gemini-3.1-pro-preview	Google	71.2%	B	68.4–73.8	12	66%	66.7%	79.2%	42.9%	86.1%	86.5%	±7.5
9	DeepSeek V4 Pro deepseek-v4-pro	DeepSeek	67.3%	C	63.2–71.3	4	82.3%	61.3%	72.9%	32.6%	81.3%	73.4%	±13.5