Skip to content
Integrated Intelligence
RUOpen Neuro

Frontier AI models

Who leads in reasoning, code and science, who is fastest and who gives the most for the money — from independent evaluations, one row per model.

Updated September 28 · Artificial Analysis data

Category leaders

57.6out of 100

Intelligence

Claude Opus 5.5

Anthropic

81.6out of 100

Coding

Claude Fable 5.1

Anthropic

96.1%correct answers

Science

GPT-6 Astra

OpenAI

61.4%correct answers

Hard exams

Claude Opus 5.5

Anthropic

291 tok/s

Speed

Gemini 3.8 Flash

Google

176index points per $1

Value

GLM 5.3 Flash

Z AI

Ranking

The Artificial Analysis composite: reasoning, knowledge, code and agentic work.

#ModelIntelligenceIntelligenceCodeSpeed$ per 1M
1Claude Opus 5.5AnthropicIn Neuro57.657.6—96$8.0
2Claude Fable 5.1AnthropicIn Neuro53.453.481.669$20
3GPT-6 AstraOpenAIIn Neuro52.752.776.962$20
4Muse Spark 1.3Meta48.148.175.8160$2.0
5GPT-6 SolOpenAIIn Neuro47.547.5—87$4.0
6Grok 4.7SpaceXAIIn Neuro46.446.4—72$3.0
7MiMo-V2.6-ProXiaomi46.346.3—41$0.54
8Qwen3.8 MaxAlibabaIn Neuro45.445.476.238$3.0
9GLM-5.3Z AI44.844.874.884$2.1
10Step 5 PreviewStepFun43.743.7—78$1.4
11Kimi K3KimiIn Neuro43.643.676.238$6.0
12GPT-5.6 TerraOpenAI42.142.176.798$4.5
13GLM 5.3 FlashZ AIIn Neuro41.841.871.557$0.24
14Gemini 3.8 FlashGoogle40.940.976.3291$1.5
15Qwen3.8 2.4T A95BAlibaba39.939.971.938$3.0

Intelligence and price

The higher and further left a dot, the more intelligence for less money.

2530354045505560$0.1$0.3$1$3$10$30Claude Opus 5.5 · 57.6 · $8.0Claude Fable 5.1 · 53.4 · $20GPT-6 Astra · 52.7 · $20Muse Spark 1.3 · 48.1 · $2.0GPT-6 Sol · 47.5 · $4.0Grok 4.7 · 46.4 · $3.0MiMo-V2.6-Pro · 46.3 · $0.54Qwen3.8 Max · 45.4 · $3.0GLM-5.3 · 44.8 · $2.1Step 5 Preview · 43.7 · $1.4Kimi K3 · 43.6 · $6.0GPT-5.6 Terra · 42.1 · $4.5GLM 5.3 Flash · 41.8 · $0.24Gemini 3.8 Flash · 40.9 · $1.5Qwen3.8 2.4T A95B · 39.9 · $3.0Qwen3.8-Flash-Next · 39.8 · $0.23DeepSeek V4.1 Flash · 39.5 · $0.53GPT-5.4 · 39.0 · $5.6Claude Sonnet 5 · 38.2 · $4.0MiMo-V2.6-Flash · 37.9 · $0.17GPT-5.6 Luna · 37.3 · $0.45DeepSeek V4 Pro 0813 · 36.0 · $2.0DeepSeek V4 Flash Vision · 34.8 · $0.66Qwen3.8 27B · 33.7 · $1.1GPT-5.3 Codex · 32.5 · $4.8Gemini 3.1 Pro Preview · 29.7 · $4.5MiniMax-M3 · 29.2 · $0.53Qwen3.6 Max Preview · 28.4 · $2.9Solar Pro 4 · 28.2 · $0.53Inkling Small · 27.8 · $0.53Grok Build 0.1 0616 · 27.2 · $1.3Qwen3.6 Plus · 27.0 · $1.1Quasar 438B · 26.7 · $0.90Apodex 1.1 · 26.4 · $0.97Gemini 3 Flash Preview · 26.3 · $1.1GPT-5.5 Instant · 26.0 · $11Claude Opus 5.5Claude Fable 5.1GPT-6 AstraMuse Spark 1.3GPT-6 SolGrok 4.7MiMo-V2.6-ProGLM 5.3 FlashQwen3.8-Flash-NextMiMo-V2.6-Flash
Horizontal: price per 1M tokens (3:1 input/output, log scale). Vertical: intelligence index. The line joins models no cheaper model beats.

How to read the ranking

One row per model

Many models ship several reasoning modes. The ranking shows each model at its best mode; «All versions» also shows earlier generations.

Benchmarks are not your task

Indices show the overall level, but the best model for your work is decided on your own tasks. In Neuro you can switch models in one click and compare.

Prices and speed change

Prices are per 1M tokens from the model's developer; speed is a median of measurements. The data refreshes daily.

Artificial Analysis methodology