Changelog

Track the latest AI model benchmarks, performance evaluations, and platform updates.

Hide API Endpoints

Changelog begins 24 Dec 2025 for image, video, and speech entries

15 May 2026

Microsoft Azure performance results now available

Kimi K2.6 (Non-reasoning)

14 May 2026

CoreWeave performance results now available

Kimi K2.6 (Non-reasoning)

SiliconFlow performance results now available

Gemma 4 31B (Reasoning)

Gemma 4 26B A4B (Non-reasoning)

Gemma 4 31B (Non-reasoning)

GMI performance results now available

Gemma 4 26B A4B (Non-reasoning)

Novita performance results now available

Gemma 4 26B A4B (Non-reasoning)

Qwen3.6 35B A3B (Reasoning)

Qwen3.6 35B A3B (Non-reasoning)

FriendliAI performance results now available

Gemma 4 31B (Non-reasoning)

Together.ai performance results now available

Gemma 4 31B (Non-reasoning)

Wafer performance results now available

Qwen3.5 397B A17B (Reasoning)

Qwen3.5 397B A17B (Non-reasoning)

13 May 2026

Novita performance results now available

Qwen3.6 27B (Reasoning)

MiMo-V2.5-Pro (Non-reasoning)

MiMo-V2.5-Pro

Nebius performance results now available

DeepSeek V4 Pro (Reasoning, High Effort)

DeepSeek V4 Pro (Non-reasoning)

GMI performance results now available

Hy3-preview (Reasoning)

Hy3-preview (Non-reasoning)

12 May 2026

New language model evaluation results available

JT-35B-Flash

Intelligence Index: 36 | AA-Omniscience: -23

Nebius performance results now available

GLM-5.1 (Reasoning)

GLM-5.1 (Non-reasoning)

DeepSeek V4 Pro (Reasoning, Max Effort)

SiliconFlow performance results now available

Qwen3.6 27B (Reasoning)

Qwen3.5 397B A17B (Reasoning)

Qwen3.5 122B A10B (Reasoning)

Qwen3.5 35B A3B (Reasoning)

Qwen3.5 27B (Reasoning)

Qwen3.6 35B A3B (Reasoning)

11 May 2026

🚀 New website feature

Artificial Analysis Coding Agent Benchmarks launch

Compare coding agents across SWE-Bench-Pro-Hard-AA, Terminal-Bench v2, and SWE-Atlas-QnA, with index, cost, speed, and token usage benchmarks.

🔔 New article published

OpenBMB launches MiniCPM-V 4.6 1.3B Instruct

New model released by OpenBMB

New language model evaluation results available

MiniCPM-V 4.6 1.3B

Intelligence Index: 13 | AA-Omniscience: -85

08 May 2026

New language model evaluation results available

Grok 4.3 (Non-reasoning)

Intelligence Index: 31 | AA-Omniscience: -32

xAI performance results now available

Grok 4.3 (Non-reasoning)

06 May 2026

DeepInfra performance results now available

MiMo-V2.5-Pro (Non-reasoning)

MiMo-V2.5-Pro

Qwen3.6 27B (Reasoning)

Qwen3.6 27B (Non-reasoning)

Lightning AI performance results now available

DeepSeek V4 Pro (Non-reasoning)

DeepSeek V4 Pro (Reasoning, Max Effort)

Baseten performance results now available

DeepSeek V4 Pro (Reasoning, High Effort)

GMI performance results now available

MiMo-V2.5-Pro (Non-reasoning)

MiMo-V2.5-Pro

MiMo-V2.5

CoreWeave performance results now available

DeepSeek V4 Flash (Non-reasoning)

05 May 2026

Microsoft Azure performance results now available

Kimi K2.6

Kimi K2.5 (Non-reasoning)

Grok 4.20 0309 v2 (Reasoning)

Kimi K2.5 (Reasoning)

Clarifai performance results now available

Nemotron 3 Nano Omni 30B A3B Reasoning

Nebius performance results now available

Nemotron 3 Nano Omni 30B A3B Reasoning

04 May 2026

New language model evaluation results available

Nemotron 3 Nano Omni 30B A3B Reasoning

Intelligence Index: 21 | AA-Omniscience: -56

SambaNova performance results now available

MiniMax-M2.7

Xiaomi performance results now available

MiMo-V2.5

FriendliAI performance results now available

Gemma 4 31B (Reasoning)

CoreWeave performance results now available

Kimi K2.6

02 May 2026

🚀 New website feature

Cache pricing now available in language model pricing

Compare cached input token prices across models and providers, with expanded blended price ratios (cache hit: input: output)

30 Apr 2026

🔔 New article published

Recent open weights model launches

Reflection on recent open weights model launches from Moonshot AI, Xiaomi, and DeepSeek

🔔 New article published

xAI launches Grok 4.3 with improved agentic performance and lower pricing

Benchmarks and Analysis of Grok 4.3

New language model evaluation results available

Hy3-preview (Non-reasoning)

Intelligence Index: 34 | AA-Omniscience: -36

New language model evaluation results available

Grok 4.3 (high)

Intelligence Index: 53 | AA-Omniscience: 18

New language model evaluation results available

Mistral Medium 3.5

Intelligence Index: 39 | AA-Omniscience: -36

New language model evaluation results available

GPT-5.5 Pro (xhigh)

SiliconFlow performance results now available

Hy3-preview (Non-reasoning)

xAI performance results now available

Grok 4.3 (high)

Mistral performance results now available

Mistral Medium 3.5

GMI performance results now available

MiniMax-M2.5

Clarifai performance results now available

gpt-oss-20B (low)

gpt-oss-20B (high)

Parasail performance results now available

DeepSeek V4 Flash (Reasoning, Max Effort)

CompactifAI performance results now available

gpt-oss-20B (high)

gpt-oss-120B (high)

DeepInfra performance results now available

Mistral Small 3.1

Cohere performance results now available

Tiny Aya Global

Kimi performance results now available

Kimi K2.6 (Non-reasoning)

Xiaomi performance results now available

MiMo-V2-Omni

MiMo-V2.5-Pro (Non-reasoning)

MiMo-V2-Omni-0327

DeepSeek performance results now available

DeepSeek V4 Pro (Non-reasoning)

DeepSeek V4 Flash (Non-reasoning)

29 Apr 2026

New language model evaluation results available

MiMo-V2.5-Pro (Non-reasoning)

Intelligence Index: 36 | AA-Omniscience: -38

New language model evaluation results available

DeepSeek V4 Flash (Non-reasoning)

Intelligence Index: 36 | AA-Omniscience: -44

New language model evaluation results available

DeepSeek V4 Pro (Non-reasoning)

Intelligence Index: 39 | AA-Omniscience: -30

New language model evaluation results available

Kimi K2.6 (Non-reasoning)

Intelligence Index: 43 | AA-Omniscience: -10

New language model evaluation results available

Granite 4.1 3B

Intelligence Index: 9 | AA-Omniscience: -77

New language model evaluation results available

Granite 4.1 30B

Intelligence Index: 15 | AA-Omniscience: -68

New language model evaluation results available

Granite 4.1 8B

Intelligence Index: 12 | AA-Omniscience: -65

GMI performance results now available

MiniMax-M2.7

OpenAI performance results now available

GPT-5.4 (low)

Parasail performance results now available

DeepSeek V4 Flash (Reasoning, High Effort)

DeepInfra performance results now available

DeepSeek V4 Flash (Reasoning, Max Effort)

CoreWeave performance results now available

Granite 4.1 8B

28 Apr 2026

DeepInfra performance results now available

DeepSeek V4 Pro (Reasoning, Max Effort)

Fireworks performance results now available

DeepSeek V4 Pro (Reasoning, Max Effort)

DigitalOcean performance results now available

MiniMax-M2.5

DeepSeek V3.2 (Non-reasoning)

Qwen3.5 397B A17B (Reasoning)

Qwen3.5 397B A17B (Non-reasoning)

DeepSeek V3.2 (Reasoning)

27 Apr 2026

New language model evaluation results available

Hy3-preview (Reasoning)

Intelligence Index: 42 | AA-Omniscience: -35

New language model evaluation results available

EXAONE 4.5 33B (Non-reasoning)

New language model evaluation results available

EXAONE 4.5 33B

Intelligence Index: 30 | AA-Omniscience: -51

New language model evaluation results available

MiMo-V2.5

Intelligence Index: 49 | AA-Omniscience: -9

Novita performance results now available

DeepSeek V4 Pro (Reasoning, Max Effort)

Qwen3.6 27B (Non-reasoning)

DeepSeek V4 Flash (Reasoning, Max Effort)

DeepSeek V4 Flash (Reasoning, High Effort)

DeepSeek V4 Pro (Reasoning, High Effort)

SiliconFlow performance results now available

Hy3-preview (Reasoning)

DeepSeek V4 Pro (Reasoning, Max Effort)

DeepSeek V4 Flash (Reasoning, Max Effort)

DeepSeek V4 Pro (Reasoning, High Effort)

DeepSeek V4 Flash (Reasoning, High Effort)

DeepInfra performance results now available

DeepSeek V4 Flash (Reasoning, High Effort)

Together.ai performance results now available

DeepSeek V4 Pro (Reasoning, Max Effort)

Fireworks performance results now available

DeepSeek V4 Pro (Reasoning, High Effort)

24 Apr 2026

🔔 New article published

DeepSeek is back among the leading open weights models with V4 Pro and V4 Flash

DeepSeek is back among the leading open weights models with the release of DeepSeek V4 Pro and V4 Flash, with V4 Pro second only to Kimi K2.6 on the Artificial Analysis Intelligence Index

New language model evaluation results available

Qwen3.6 27B (Non-reasoning)

Intelligence Index: 37 | AA-Omniscience: -53

DeepInfra performance results now available

Kimi K2.6

DeepSeek V4 Pro (Reasoning, High Effort)

SiliconFlow performance results now available

Kimi K2.6

Clarifai performance results now available

Kimi K2.6

Lightning AI performance results now available

DeepSeek V4 Pro (Reasoning, High Effort)

Together.ai performance results now available

DeepSeek V4 Pro (Reasoning, High Effort)

Alibaba Cloud performance results now available

Qwen3.6 27B (Non-reasoning)

23 Apr 2026

🔔 New article published

OpenAI's GPT-5.5 is the new leading AI model

Benchmarks and Analysis of GPT-5.5

New language model evaluation results available

DeepSeek V4 Pro (Reasoning, Max Effort)

Intelligence Index: 52 | AA-Omniscience: -10

New language model evaluation results available

DeepSeek V4 Pro (Reasoning, High Effort)

Intelligence Index: 50 | AA-Omniscience: -10

New language model evaluation results available

DeepSeek V4 Flash (Reasoning, Max Effort)

Intelligence Index: 47 | AA-Omniscience: -23

New language model evaluation results available

DeepSeek V4 Flash (Reasoning, High Effort)

Intelligence Index: 46 | AA-Omniscience: -22

New language model evaluation results available

GPT-5.5 (low)

Intelligence Index: 51 | AA-Omniscience: 15

New language model evaluation results available

GPT-5.5 (high)

Intelligence Index: 59 | AA-Omniscience: 18

New language model evaluation results available

GPT-5.5 (Non-reasoning)

Intelligence Index: 41 | AA-Omniscience: -5

New language model evaluation results available

GPT-5.5 (medium)

Intelligence Index: 57 | AA-Omniscience: 17

New language model evaluation results available

GPT-5.5 (xhigh)

Intelligence Index: 60 | AA-Omniscience: 20

New language model evaluation results available

Ling-2.6-1T

Intelligence Index: 34 | AA-Omniscience: -51

New language model evaluation results available

Qwen3.6 27B (Reasoning)

Intelligence Index: 46 | AA-Omniscience: -20

DeepSeek performance results now available

DeepSeek V4 Flash (Reasoning, Max Effort)

DeepSeek V4 Pro (Reasoning, High Effort)

DeepSeek V4 Flash (Reasoning, High Effort)

DeepSeek V4 Pro (Reasoning, Max Effort)

OpenAI performance results now available

GPT-5.5 (low)

GPT-5.5 (Non-reasoning)

GPT-5.5 (medium)

GPT-5.5 (high)

GPT-5.5 (xhigh)

Together.ai performance results now available

Kimi K2.6

InclusionAI performance results now available

Ling-2.6-1T

Alibaba Cloud performance results now available

Qwen3.6 27B (Reasoning)

22 Apr 2026

New language model evaluation results available

MiMo-V2.5-Pro

Intelligence Index: 54 | AA-Omniscience: 4

Xiaomi performance results now available

MiMo-V2.5-Pro

21 Apr 2026

New language model evaluation results available

Claude Opus 4.7 (Non-reasoning, High Effort)

Intelligence Index: 52 | AA-Omniscience: 14

New language model evaluation results available

Qwen3.6 35B A3B (Non-reasoning)

Intelligence Index: 32 | AA-Omniscience: -60

New language model evaluation results available

Ling 2.6 Flash

Intelligence Index: 26 | AA-Omniscience: -66

Novita performance results now available

Ling 2.6 Flash

Kimi K2.6

Cloudflare performance results now available

Gemma 4 26B A4B (Reasoning)

Kimi K2.6

Amazon Bedrock performance results now available

Claude Opus 4.7 (Non-reasoning, High Effort)

Anthropic performance results now available

Claude Opus 4.7 (Non-reasoning, High Effort)

Clarifai performance results now available

Qwen3.6 35B A3B (Non-reasoning)

Qwen3.6 35B A3B (Reasoning)

Parasail performance results now available

Qwen3.6 35B A3B (Non-reasoning)

Qwen3.6 35B A3B (Reasoning)

Kimi K2.6

DeepInfra performance results now available

Qwen3.6 35B A3B (Non-reasoning)

Qwen3.6 35B A3B (Reasoning)

Alibaba Cloud performance results now available

Qwen3.6 35B A3B (Non-reasoning)

Qwen3.6 Max Preview

Fireworks performance results now available

Kimi K2.6

Kimi performance results now available

Kimi K2.6

20 Apr 2026

🔔 New article published

Kimi K2.6: The new leading open weights model

Benchmarks and Analysis of Kimi K2.6

New language model evaluation results available

Kimi K2.6

Intelligence Index: 54 | AA-Omniscience: 6

19 Apr 2026

New language model evaluation results available

Qwen3.6 Max Preview

Intelligence Index: 52 | AA-Omniscience: 10

17 Apr 2026

🔔 New article published

Opus 4.7: Everything you need to know

Benchmarks and Analysis of Opus 4.7

New language model evaluation results available

Qwen3.6 35B A3B (Reasoning)

Intelligence Index: 43 | AA-Omniscience: -21

Alibaba Cloud performance results now available

Qwen3.6 35B A3B (Reasoning)

Anthropic performance results now available

Claude Opus 4.7 (Adaptive Reasoning, Max Effort)

Google performance results now available

Claude Opus 4.7 (Adaptive Reasoning, Max Effort)

Microsoft Azure performance results now available

Claude Opus 4.7 (Adaptive Reasoning, Max Effort)

Amazon Bedrock performance results now available

Claude Opus 4.7 (Adaptive Reasoning, Max Effort)

Together.ai performance results now available

Qwen3.5 397B A17B (Non-reasoning)

16 Apr 2026

New language model evaluation results available

Claude Opus 4.7 (Adaptive Reasoning, Max Effort)

Intelligence Index: 57 | AA-Omniscience: 26

New model in Text to Image Leaderboard

ImagineArt 2.0

Text to Image Elo: 1180

15 Apr 2026

New language model evaluation results available

JT-MINI

Intelligence Index: 25 | AA-Omniscience: -65

DeepInfra performance results now available

DeepSeek V3.2 (Reasoning)

14 Apr 2026

DeepInfra performance results now available

Gemma 4 31B (Reasoning)

Fireworks performance results now available

MiniMax-M2.7

Together.ai performance results now available

MiniMax-M2.7

Clarifai performance results now available

Gemma 4 26B A4B (Reasoning)

CoreWeave performance results now available

Gemma 4 31B (Reasoning)

13 Apr 2026

🔔 New article published

Sub-32B Open Weights

Benchmark and analyses with focus on Qwen3.5 and Gemma 4 families

New language model evaluation results available

Step 3.5 Flash 2603

Intelligence Index: 38 | AA-Omniscience: -44

DeepInfra performance results now available

Gemma 4 26B A4B (Reasoning)

Together.ai performance results now available

Gemma 4 31B (Reasoning)

GLM-5.1 (Reasoning)

CoreWeave performance results now available

GLM-5.1 (Reasoning)

GMI performance results now available

Gemma 4 26B A4B (Reasoning)

StepFun performance results now available

Step 3.5 Flash 2603

10 Apr 2026

New model added to Speech Arena Leaderboard

MiMo-V2-TTS

Text to Speech Elo: 1011

09 Apr 2026

🔔 New article published

A new look for Artificial Analysis

Same mission, new look

New language model evaluation results available

GLM-5.1 (Non-reasoning)

Intelligence Index: 44 | AA-Omniscience: -21

SiliconFlow performance results now available

GLM-5.1 (Non-reasoning)

Novita performance results now available

GLM-5.1 (Non-reasoning)

Parasail performance results now available

GLM-5.1 (Non-reasoning)

DeepInfra performance results now available

GLM-5.1 (Non-reasoning)

FriendliAI performance results now available

GLM-5.1 (Non-reasoning)

08 Apr 2026

🔔 New article published

Muse Spark: Meta is back in the AI race

Benchmarks and Analysis of Muse Spark

New language model evaluation results available

Muse Spark

Intelligence Index: 52 | AA-Omniscience: 4

New language model evaluation results available

Trinity Large Thinking

Intelligence Index: 32 | AA-Omniscience: -44

New language model evaluation results available

Grok 4.20 0309 v2 (Reasoning)

Intelligence Index: 49 | AA-Omniscience: 15

New language model evaluation results available

Grok 4.20 0309 v2 (Non-reasoning)

Intelligence Index: 29 | AA-Omniscience: -45

New language model evaluation results available

Qwen3.6 Plus

Intelligence Index: 50 | AA-Omniscience: 3

Arcee AI performance results now available

Trinity Large Thinking

Parasail performance results now available

Trinity Large Thinking

xAI performance results now available

Grok 4.20 0309 v2 (Reasoning)

Grok 4.20 0309 v2 (Non-reasoning)

Alibaba Cloud performance results now available

Qwen3.6 Plus

07 Apr 2026

New language model evaluation results available

MiMo-V2-Omni-0327

Intelligence Index: 45 | AA-Omniscience: -14

New language model evaluation results available

GLM-5.1 (Reasoning)

Intelligence Index: 51 | AA-Omniscience: 2

New language model evaluation results available

Qwen3.5 Omni Flash

Intelligence Index: 26 | AA-Omniscience: -66

SiliconFlow performance results now available

GLM-5.1 (Reasoning)

FriendliAI performance results now available

GLM-5.1 (Reasoning)

DeepInfra performance results now available

GLM-5.1 (Reasoning)

Fireworks performance results now available

GLM-5.1 (Reasoning)

GMI performance results now available

GLM-5.1 (Reasoning)

Novita performance results now available

GLM-5.1 (Reasoning)

Parasail performance results now available

GLM-5.1 (Reasoning)

Alibaba Cloud performance results now available

Qwen3.5 Omni Flash

Amazon Bedrock performance results now available

Nova 2.0 Lite (high)

Nova 2.0 Omni (medium)

Nova 2.0 Lite (high)

New model added to Speech Arena Leaderboard

MAI-Voice-1

Text to Speech Elo: 1023

06 Apr 2026

🔔 New article published

Gemma 4: Google on the American open weights frontier

Benchmarks and Analysis of Gemma 4 models

New language model evaluation results available

Solar Pro 3

Intelligence Index: 26 | AA-Omniscience: -54

New language model evaluation results available

Gemma 4 26B A4B (Non-reasoning)

Intelligence Index: 27 | AA-Omniscience: -62

New language model evaluation results available

Gemma 4 31B (Non-reasoning)

Intelligence Index: 32 | AA-Omniscience: -51

New language model evaluation results available

Gemma 4 E4B (Non-reasoning)

Intelligence Index: 15 | AA-Omniscience: -42

New language model evaluation results available

Gemma 4 E2B (Non-reasoning)

Intelligence Index: 12 | AA-Omniscience: -61

GMI performance results now available

Gemma 4 31B (Reasoning)

Lightning AI performance results now available

MiniMax-M2.5

03 Apr 2026

New language model evaluation results available

Gemma 4 E4B (Reasoning)

Intelligence Index: 19 | AA-Omniscience: -20

New language model evaluation results available

Gemma 4 E2B (Reasoning)

Intelligence Index: 15 | AA-Omniscience: -24

Google performance results now available

Gemma 4 31B (Reasoning)

Gemma 4 26B A4B (Reasoning)

Parasail performance results now available

Gemma 4 26B A4B (Reasoning)

Gemma 4 31B (Reasoning)

Novita performance results now available

Gemma 4 26B A4B (Reasoning)

Gemma 4 31B (Reasoning)

Lightning AI performance results now available

GLM-5 (Reasoning)

02 Apr 2026

🔔 New article published

Sarvam 105B and Sarvam 30B: India enters the open-weights race

Benchmarks and Analysis of Sarvam 105B and Sarvam 30B

🔔 New article published

MAI-Transcribe-1: Everything you need to know

Benchmarks and Analysis of MAI-Transcribe-1

New language model evaluation results available

Qwen3.5 Omni Plus

Intelligence Index: 39 | AA-Omniscience: -12

New language model evaluation results available

Gemma 4 26B A4B (Reasoning)

Intelligence Index: 31 | AA-Omniscience: -48

New language model evaluation results available

Gemma 4 31B (Reasoning)

Intelligence Index: 39 | AA-Omniscience: -45

Alibaba Cloud performance results now available

Qwen3.5 Omni Plus

New Speech to Text model evaluation results available

MAI-Transcribe-1, Azure

01 Apr 2026

New language model evaluation results available

GLM 5V Turbo (Reasoning)

Intelligence Index: 43 | AA-Omniscience: -19

New model added to Speech Arena Leaderboard

Voxtral TTS

Text to Speech Elo: 1056

31 Mar 2026

New model added to Speech Arena Leaderboard

Magpie-Multilingual 357M (Feb 2026)

Text to Speech Elo: 1063

30 Mar 2026

New language model evaluation results available

Nemotron Cascade 2 30B A3B

Intelligence Index: 28 | AA-Omniscience: -52

New language model evaluation results available

KAT Coder Pro V2

Intelligence Index: 44 | AA-Omniscience: -22

DeepInfra performance results now available

Qwen3.5 0.8B (Non-reasoning)

Qwen3.5 2B (Non-reasoning)

Qwen3.5 4B (Non-reasoning)

Qwen3.5 27B (Non-reasoning)

Qwen3.5 35B A3B (Non-reasoning)

Qwen3.5 122B A10B (Non-reasoning)

Qwen3.5 397B A17B (Non-reasoning)

Qwen3.5 0.8B (Reasoning)

Qwen3.5 2B (Reasoning)

Qwen3.5 4B (Reasoning)

Qwen3.5 27B (Reasoning)

Qwen3.5 35B A3B (Reasoning)

Qwen3.5 397B A17B (Reasoning)

Qwen3.5 122B A10B (Reasoning)

StreamLake performance results now available

KAT Coder Pro V2

26 Mar 2026

New language model evaluation results available

Gemini 3 Deep Think

New model added to Speech Arena Leaderboard

Step TTS 2 (Mar 2026)

Text to Speech Elo: 1148

New model added to Speech Arena Leaderboard

Step Audio EditX (Mar 2026)

Text to Speech Elo: 1104

New Speech to Text model evaluation results available

Cohere Transcribe

25 Mar 2026

🔔 New article published

MiniMax M2.7: Everything you need to know

Benchmarks and Analysis of MiniMax M2.7

Xiaomi performance results now available

MiMo-V2-Pro

New model added to Speech Arena Leaderboard

SIMBA 1.6

Text to Speech Elo: 1060

24 Mar 2026

New language model evaluation results available

NVIDIA Nemotron 3 Nano 4B

Intelligence Index: 15 | AA-Omniscience: -72

New language model evaluation results available

GLM-5-Turbo

Intelligence Index: 47 | AA-Omniscience: -15

New model added to Speech Arena Leaderboard

Fish Audio S2 Pro

Text to Speech Elo: 1128

21 Mar 2026

New language model evaluation results available

Apertus 8B Instruct

Intelligence Index: 6 | AA-Omniscience: -75

New language model evaluation results available

Apertus 70B Instruct

Intelligence Index: 8 | AA-Omniscience: -55

Public AI performance results now available

Apertus 70B Instruct

Apertus 8B Instruct

20 Mar 2026

🔔 New article published

MiMo-V2-Pro: Everything you need to know

Benchmarks and Analysis of MiMo-V2-Pro

New language model evaluation results available

Nanbeige4.1-3B

Intelligence Index: 16 | AA-Omniscience: -42

New language model evaluation results available

MiMo-V2-Omni

Intelligence Index: 43 | AA-Omniscience: -17

19 Mar 2026

New language model evaluation results available

GPT-5.4 nano (Non-Reasoning)

Intelligence Index: 24 | AA-Omniscience: -39

New language model evaluation results available

GPT-5.4 nano (medium)

Intelligence Index: 38 | AA-Omniscience: -19

OpenAI performance results now available

GPT-5.4 mini (medium)

GPT-5.4 nano (medium)

GPT-5.4 nano (Non-Reasoning)

GPT-5.4 mini (xhigh)

GPT-5.4 mini (Non-Reasoning)

GPT-5.4 nano (xhigh)

Microsoft Azure performance results now available

GPT-5.4 (xhigh)

GPT-5.4 mini (Non-Reasoning)

GPT-5.4 mini (medium)

GPT-5.4 mini (xhigh)

18 Mar 2026

New language model evaluation results available

GPT-5.4 mini (Non-Reasoning)

Intelligence Index: 23 | AA-Omniscience: -46

New language model evaluation results available

GPT-5.4 mini (medium)

Intelligence Index: 38 | AA-Omniscience: -20

New language model evaluation results available

GPT-5.4 mini (xhigh)

Intelligence Index: 49 | AA-Omniscience: -19

New language model evaluation results available

GPT-5.4 nano (xhigh)

Intelligence Index: 44 | AA-Omniscience: -30

New language model evaluation results available

Sarvam 105B (high)

Intelligence Index: 18 | AA-Omniscience: -60

New language model evaluation results available

Sarvam 30B (high)

Intelligence Index: 12 | AA-Omniscience: -72

New language model evaluation results available

GPT-5.4 (Non-reasoning)

Intelligence Index: 35 | AA-Omniscience: -16

New language model evaluation results available

MiniMax-M2.7

Intelligence Index: 50 | AA-Omniscience: 1

New language model evaluation results available

Mistral Small 4 (Non-reasoning)

Intelligence Index: 19 | AA-Omniscience: -48

New language model evaluation results available

Mistral Small 4 (Reasoning)

Intelligence Index: 28 | AA-Omniscience: -30

Sarvam performance results now available

Sarvam 30B (high)

Sarvam 105B (high)

Nebius performance results now available

MiniMax-M2.5

MiniMax performance results now available

MiniMax-M2.7

Novita performance results now available

MiniMax-M2.7

Mistral performance results now available

Mistral Small 4 (Non-reasoning)

Mistral Small 4 (Reasoning)

17 Mar 2026

New language model evaluation results available

MiMo-V2-Pro

Intelligence Index: 49 | AA-Omniscience: 5

Nebius performance results now available

gpt-oss-120B (high)

Qwen3 235B A22B 2507 (Reasoning)

Qwen3.5 397B A17B (Non-reasoning)

Qwen3 Next 80B A3B (Reasoning)

Qwen3.5 397B A17B (Reasoning)

Qwen3.5 397B A17B (Non-reasoning)

DeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2 (Reasoning)

16 Mar 2026

Minor scoring update — no change to Intelligence Index

GPT-5.4 (xhigh): CritPT score updated from 20.0% to 23.4%

What's changed

CritPT score updated for GPT-5.4 (xhigh): We observed higher variance in individual CritPT scores for GPT-5.4 (xhigh) and ran additional evaluation runs to confirm. The CritPT score has been updated from 20.0% to 23.4%.
Intelligence Index unchanged: This does not affect the overall Intelligence Index score for GPT-5.4 (xhigh), as the change is within the Intelligence Index margin of error. All other benchmark scores remain unchanged.

Clarifai performance results now available

Qwen3.5 397B A17B (Reasoning)

13 Mar 2026

🔔 New article published

NVIDIA Nemotron 3 VoiceChat: Leading the Open Weights Frontier of Conversational Dynamics vs. Speech Reasoning

Independent analysis and benchmarks of NVIDIA Nemotron 3 VoiceChat

Eigen AI performance results now available

DeepSeek V3.2 (Reasoning)

DeepSeek V3.2 (Non-reasoning)

GMI performance results now available

Qwen3.5 397B A17B (Reasoning)

Qwen3.5 27B (Reasoning)

Qwen3.5 35B A3B (Reasoning)

Qwen3.5 122B A10B (Reasoning)

FriendliAI performance results now available

Kimi K2.5 (Non-reasoning)

Kimi K2.5 (Reasoning)

Amazon Bedrock performance results now available

Qwen3 Coder Next

12 Mar 2026

Nebius performance results now available

Kimi K2.5 (Reasoning)

GLM-5 (Non-reasoning)

GLM-5 (Reasoning)

11 Mar 2026

🔔 New article published

NVIDIA Nemotron 3 Super: The new leader in open, efficient intelligence

Independent analysis and benchmarks of NVIDIA Nemotron 3 Super

New language model evaluation results available

Grok 4.20 0309 (Non-reasoning)

Intelligence Index: 30 | AA-Omniscience: -43

New language model evaluation results available

Grok 4.20 0309 (Reasoning)

Intelligence Index: 48 | AA-Omniscience: 13

New language model evaluation results available

Qwen3.5 0.8B (Non-reasoning)

Intelligence Index: 10 | AA-Omniscience: -89

New language model evaluation results available

NVIDIA Nemotron 3 Super 120B A12B (Reasoning)

Intelligence Index: 36 | AA-Omniscience: -42

New language model evaluation results available

LongCat Flash Lite

Intelligence Index: 24 | AA-Omniscience: -70

xAI performance results now available

Grok 4.20 0309 (Reasoning)

Grok 4.20 0309 (Non-reasoning)

CoreWeave performance results now available

NVIDIA Nemotron 3 Super 120B A12B (Reasoning)

Nebius performance results now available

NVIDIA Nemotron 3 Super 120B A12B (Reasoning)

Baseten performance results now available

NVIDIA Nemotron 3 Super 120B A12B (Reasoning)

DeepInfra performance results now available

NVIDIA Nemotron 3 Super 120B A12B (Reasoning)

LongCat performance results now available

LongCat Flash Lite

10 Mar 2026

Eigen AI performance results now available

Qwen3.5 397B A17B (Reasoning)

MiniMax-M2.5

Qwen3 8B (Non-reasoning)

Qwen3 30B A3B (Reasoning)

Qwen3 VL 30B A3B (Reasoning)

Qwen3 VL 30B A3B Instruct

Qwen3 8B (Reasoning)

Qwen3 30B A3B (Non-reasoning)

Qwen3.5 397B A17B (Non-reasoning)

Clarifai performance results now available

Kimi K2.5 (Reasoning)

New model in Text to Video Leaderboard

LTX-2.3 Fast

Text to Video Elo: 1124

New model in Text to Video Leaderboard

LTX-2.3 Pro

Text to Video Elo: 1110

New model in Image to Video Leaderboard

LTX-2.3 Fast

Image to Video Elo: 1158

New model in Image to Video Leaderboard

LTX-2.3 Pro

Image to Video Elo: 1155

09 Mar 2026

New language model evaluation results available

Qwen3.5 2B (Non-reasoning)

Intelligence Index: 15 | AA-Omniscience: -83

New language model evaluation results available

Qwen3.5 4B (Non-reasoning)

Intelligence Index: 23 | AA-Omniscience: -75

New language model evaluation results available

Qwen3.5 9B (Non-reasoning)

Intelligence Index: 27 | AA-Omniscience: -71

Lightning AI performance results now available

NVIDIA Nemotron 3 Super 120B A12B (Reasoning)

FriendliAI performance results now available

DeepSeek V3.2 (Non-reasoning)

Together.ai performance results now available

Qwen3.5 9B (Reasoning)

07 Mar 2026

CoreWeave performance results now available

Qwen3 30B A3B 2507 Instruct

06 Mar 2026

New language model evaluation results available

GPT-5.4 Pro (xhigh)

OpenAI performance results now available

GPT-5.4 (Non-reasoning)

GPT-5.4 Pro (xhigh)

SiliconFlow performance results now available

Qwen2.5 Instruct 72B

Step 3.5 Flash

FriendliAI performance results now available

MiniMax-M2.5

MiniMax-M2.1

GLM-5 (Reasoning)

Qwen3 235B A22B 2507 Instruct

CoreWeave performance results now available

MiniMax-M2.5

Kimi K2.5 (Reasoning)

GLM-5 (Reasoning)

05 Mar 2026

🔔 New article published

Qwen3.5 small models: Everything you need to know

Independent analysis and benchmarks of Qwen 3.5 small models

OpenAI performance results now available

GPT-5.4 (xhigh)

Baseten performance results now available

GLM-5 (Reasoning)

GLM-5 (Non-reasoning)

CoreWeave performance results now available

Llama 3.1 Instruct 8B

Qwen3 235B A22B 2507 Instruct

Qwen3 Coder 480B A35B Instruct

DeepSeek V3.1 (Non-reasoning)

gpt-oss-20B (high)

gpt-oss-120B (high)

Qwen3 235B A22B 2507 (Reasoning)

Llama 3.3 Instruct 70B

Amazon Bedrock performance results now available

GLM-4.7 (Non-reasoning)

GLM-4.7-Flash (Non-reasoning)

DeepSeek V3.2 (Reasoning)

DeepSeek V3.2 (Non-reasoning)

GLM-4.7 (Reasoning)

GLM-4.7-Flash (Reasoning)

04 Mar 2026

New language model evaluation results available

Qwen3.5 2B (Reasoning)

Intelligence Index: 16 | AA-Omniscience: -47

New language model evaluation results available

Qwen3.5 0.8B (Reasoning)

Intelligence Index: 11 | AA-Omniscience: -35

New language model evaluation results available

Qwen3.5 4B (Reasoning)

Intelligence Index: 27 | AA-Omniscience: -57

New language model evaluation results available

Qwen3.5 9B (Reasoning)

Intelligence Index: 32 | AA-Omniscience: -53

Amazon Bedrock performance results now available

Kimi K2.5 (Reasoning)

Nebius performance results now available

Kimi K2.5 (Non-reasoning)

Gemma 3 27B Instruct

03 Mar 2026

New language model evaluation results available

Gemini 3.1 Flash-Lite Preview

Intelligence Index: 34 | AA-Omniscience: -16

Google performance results now available

Gemini 3.1 Flash-Lite Preview

Nebius performance results now available

gpt-oss-120B (high)

02 Mar 2026

New language model evaluation results available

Qwen3.5 122B A10B (Non-reasoning)

Intelligence Index: 36 | AA-Omniscience: -54

New language model evaluation results available

Qwen3.5 27B (Non-reasoning)

Intelligence Index: 37 | AA-Omniscience: -47

New language model evaluation results available

Qwen3.5 35B A3B (Non-reasoning)

Intelligence Index: 31 | AA-Omniscience: -62

Alibaba Cloud performance results now available

Qwen3.5 27B (Non-reasoning)

Qwen3.5 35B A3B (Non-reasoning)

Qwen3.5 122B A10B (Non-reasoning)

Lightning AI performance results now available

Kimi K2.5 (Non-reasoning)

Kimi K2.5 (Reasoning)

New model in Text to Video Leaderboard

P-Video

Text to Video Elo: 1067

New model in Image to Video Leaderboard

P-Video

Image to Video Elo: 1126

01 Mar 2026

Artificial Analysis performance benchmarking methodology

Language model performance benchmarking methodology update (v2.2.0): Expanded prompt set and new default prompt length

What's changed

Refreshed prompt bank: We expanded and diversified prompts across more task types to better reflect real-world usage. This is important because techniques like speculative decoding mean that we see variation in output speeds depending on the type of output.
New default workload: Site-default speed now uses 10k input tokens (was 1k)and we have deprecated performance measurements for 100 token input workloads. 1k, 10k, and 100k remain available via Prompt Options, and the Output Speed by Input Token Count chart is the quickest way to compare workload shapes.

Server location, testing frequency, and median reporting are unchanged.