Skip to content
PAY.NEWS LEADERBOARD

AI model leaderboard

Evaluation sources How ranks are computed

Factual questions and graduate-level science — knowledge and answer accuracy.

2 evaluations·1 operators·10/17 20:05 updated

Knowledge boardTOP 30

Ranked by the shared evidence of multiple public evaluations

Benchmark aggregation, not a payment-provider rating. Sponsorship, payment and affiliations must not affect results. Editorial standards

The top 30 models, sortable by column
Model
01gemini-4-argonGoogleEvidence-sensitive · 2 evaluationsOfficial —97.4
02GPT-6 AstraOpenAIEvidence-sensitive · 2 evaluationsOfficial $10 / $5096.9
03gpt-6.1-sol—Evidence-sensitive · 2 evaluationsOfficial —95.4
04Gemini 3.1 Pro PreviewGoogleEvidence-sensitive · 2 evaluationsOfficial $2 / $1294.5
05Claude Opus 5.5AnthropicEvidence-sensitive · 2 evaluationsOfficial $4 / $2091.4
06Gemini 3.8 FlashGoogleEvidence-sensitive · 2 evaluationsOfficial $0.75 / $3.7590.2
07Gemini 3.7 FlashGoogleEvidence-sensitive · 2 evaluationsOfficial $0.75 / $3.7588.9
08GPT-5.6 SolOpenAIEvidence-sensitive · 2 evaluationsOfficial $4 / $2088.8
09Claude Fable 5AnthropicEvidence-sensitive · 2 evaluationsOfficial $10 / $5088.4
10Gemini 3.6 FlashGoogleEvidence-sensitive · 2 evaluationsOfficial $0.75 / $3.7586.2
11Gemini 3.5 FlashGoogleEvidence-sensitive · 2 evaluationsOfficial $1.5 / $985.1
12Gemini 3 Flash PreviewGoogleEvidence-sensitive · 2 evaluationsOfficial $0.5 / $384.3
13GPT-5.5OpenAIEvidence-sensitive · 2 evaluationsOfficial $5 / $3080.4
14GPT-6 SolOpenAIEvidence-sensitive · 2 evaluationsOfficial $2 / $1079.0
15Claude Opus 5AnthropicEvidence-sensitive · 2 evaluationsOfficial $5 / $2577.0
16Grok 4.7xAIEvidence-sensitive · 2 evaluationsOfficial $2 / $670.4
17Qwen3.7 MaxAlibabaEvidence-sensitive · 2 evaluationsOfficial $1.79 / $5.3868.2
18Muse-Spark-Open-816B-A42BMeta AIEvidence-sensitive · 2 evaluationsOfficial —68.1
19DeepSeek V4 Pro 0813DeepSeekEvidence-sensitive · 2 evaluationsOfficial $1.35 / $4.0464.8
20Claude Opus 4.8AnthropicEvidence-sensitive · 2 evaluationsOfficial $5 / $2564.3
21Kimi K3Moonshot AIEvidence-sensitive · 2 evaluationsOfficial $2.99 / $14.9661.5
22Qwen3.6 Max PreviewAlibabaEvidence-sensitive · 2 evaluationsOfficial $1.35 / $8.0861.4
23Claude Opus 4.7AnthropicEvidence-sensitive · 2 evaluationsOfficial $5 / $2559.9
24Grok 4.6xAIEvidence-sensitive · 2 evaluationsOfficial $2 / $657.8
25GPT-5 2025-08-07OpenAIEvidence-sensitive · 2 evaluationsOfficial $1.25 / $1057.2
26Grok 4.5xAIEvidence-sensitive · 2 evaluationsOfficial $2 / $656.6
27Qwen3.8 Max 0902AlibabaEvidence-sensitive · 2 evaluationsOfficial $1.79 / $5.3852.9
28GPT-5.1 2025-11-13OpenAIEvidence-sensitive · 2 evaluationsOfficial $1.25 / $1052.3
29Qwen3 Max 2025-09-23AlibabaEvidence-sensitive · 2 evaluationsOfficial $0.897 / $3.5951.6
30claude-sonnet-5.5anthropicEvidence-sensitive · 2 evaluationsOfficial —51.4

Each board shows at most 30 models

The consensus index is not an accuracy rate; ties keep the order the shared evidence establishes.

How to read this board

The knowledge board currently uses two Epoch evaluations, both from the same organisation.

See how it is computed →

About prices

API prices are shown in USD per million tokens by default. Vendor list prices and the DogRouter reference sit in separate columns; the latter is the relay price of a Dogpay sister service, carries a verification date and a link to its price page, is never shown as a discount, and does not affect the ranking. Cached price is the input rate on a cache hit; cache writes, storage and subscriptions are billed separately. CNY list prices are converted at the 2026-10-16 rate (欧洲央行).