Skip to content
LMC Feed-Models, Papers, Benchmarks. Zero Fluff.Live

Weekly AI Performance Report

Week of September 22, 2026

46 models improved, 249 declined, and 5 unchanged this week across 300 tracked models.

1

Executive Summary

Biggest Gainer

Claude Opus 5.5

+426 ranks

Currently ranked #5

Biggest Loser

Ling 3.0 Flash Sante (free)

-20 ranks

Currently ranked #261

New Entries

10

new models entered rankings this week

2

Weekly Movers

Top 10 models with the largest absolute rank changes over the past 7 days.

ModelScore7d ChangeRank
Claude Opus 5.5Anthropic95.1+426#5
Grok 4.7xAI88.8+388#43
MiMo-V2.6-ProXiaomi76.2+314#117
Claude Opus 5Anthropic95.1+272#6
Claude Sonnet 5Anthropic85.2+231#63
gpt-oss-20b (batch)OpenAI57.4+211#220
GPT-6 Luna ProOpenAI40.0+197#234
GPT-6 Luna Pro (batch)OpenAI40.0+196#235
GPT-6 LunaOpenAI40.0+195#236
GPT-6 Luna (batch)OpenAI40.0+194#237
3

Top Gainers

Models with the largest positive rank improvements this week.

1
Claude Opus 5.5Anthropic
+426
2
Grok 4.7xAI
+388
3
MiMo-V2.6-ProXiaomi
+314
4
Claude Opus 5Anthropic
+272
5
Claude Sonnet 5Anthropic
+231
6
gpt-oss-20b (batch)OpenAI
+211
7
GPT-6 Luna ProOpenAI
+197
8
GPT-6 Luna Pro (batch)OpenAI
+196
9
GPT-6 LunaOpenAI
+195
10
GPT-6 Luna (batch)OpenAI
+194
4

Top Losers

Models with the largest negative rank drops this week.

1
Ling 3.0 Flash Sante (free)inclusionai
-20
2
Ternary Bonsai 2 27Bprism-ml
-19
3
Paretounbiased
-19
4
DeepSeek Pro Latest~deepseek
-19
5
DeepSeek Flash Latest~deepseek
-19
6
Schematron V2 Turboinference-net
-19
7
Schematron V2 Smallinference-net
-19
8
GPT Astra Latest~openai
-19
9
GPT Sol Latest~openai
-19
10
GPT Terra Latest~openai
-19
5

New Models

10 new models entered the rankings this week.

ModelScoreRank
Claude Opus 5.5Anthropic95.1#5
Grok 4.7xAI88.8#43
MiMo-V2.6-ProXiaomi76.2#117
gpt-oss-20b (batch)OpenAI57.4#220
GPT-6 Luna ProOpenAI40.0#234
GPT-6 Luna Pro (batch)OpenAI40.0#235
GPT-6 LunaOpenAI40.0#236
GPT-6 Luna (batch)OpenAI40.0#237
GPT-6 Sol ProOpenAI40.0#238
GPT-6 Sol Pro (batch)OpenAI40.0#239
6

Watch List

Models in a fragile state that might degrade further. Monitor closely.

ModelScore7d Change
Claude Opus 5Anthropic
95.1+272
GPT-5.5 ProOpenAI
92.7+8
GPT-5.3-CodexOpenAI
90.5+18
GPT-5.2 ChatOpenAI
90.5+80
Grok 4.5xAI
88.8+56
Grok 4.3xAI
88.8+30
Grok 4.20xAI
88.8-8
GPT-5 ProOpenAI
88.7-10
GPT-5 Pro (batch)OpenAI
88.7-10
GPT-5OpenAI
88.7-10
GPT-5 (batch)OpenAI
88.7-10
Gemini 3 Flash PreviewGoogle
88.4-10
Gemini 3 Flash Preview (batch)Google
88.4-10
Grok 4.20 Multi-AgentxAI
87.9-9
GPT-5.1 (batch)OpenAI
87.8-7
GPT-5.1-Codex-MiniOpenAI
87.8-6
o3 ProOpenAI
86.7-6
o3OpenAI
86.7-6
o3 (batch)OpenAI
86.7-6
Claude Sonnet 5Anthropic
85.2+231
Claude Sonnet 4.6Anthropic
85.2-7
Claude Sonnet 4.6 (batch)Anthropic
85.2-7
Claude Opus 4.5Anthropic
85.1-7
Claude Opus 4.5 (batch)Anthropic
85.1-7
Gemini 2.5 ProGoogle
83.5-7
Gemini 2.5 Pro (batch)Google
83.5-7
Gemini 2.5 Pro Preview 06-05Google
83.5-7
DeepSeek V3.2DeepSeek
83.4-7
Claude Sonnet 4.5Anthropic
82.4-7
Claude Sonnet 4.5 (batch)Anthropic
82.4-7
GPT-6 AstraOpenAI
81.8-6
GPT-6 Astra (batch)OpenAI
81.8-6
GPT-6 Astra ProOpenAI
81.8-6
GPT-6 Astra Pro (batch)OpenAI
81.8-6
o4 MiniOpenAI
81.4-6
o4 Mini (batch)OpenAI
81.4-6
Gemini 3.8 FlashGoogle
81.0-6
Gemini 3.8 Flash (batch)Google
81.0-6
Muse Spark 1.3meta
80.9+161
Muse Spark 1.2meta
80.9+189
Muse Spark 1.1meta
80.9-7
Gemma 4 31BGoogle
80.5-7
Gemma 4 31B (free)Google
80.5-7
Qwen3.5 397B A17BAlibaba
79.4-7
R1 0528DeepSeek
79.4-7
GPT-5.4 NanoOpenAI
79.3-7
GPT-5.4 Nano (batch)OpenAI
79.3-7
GPT-5.4 MiniOpenAI
79.3-7
GPT-5.4 Mini (batch)OpenAI
79.3-7
Gemini 3.1 Flash Lite PreviewGoogle
79.3-7
Gemini 2.5 Flash LiteGoogle
79.1-7
Gemini 2.5 Flash Lite (batch)Google
79.1-7
Gemini 2.5 FlashGoogle
79.1-7
Gemini 2.5 Flash (batch)Google
79.1-7
Gemini 3.5 FlashGoogle
79.0-7
Gemini 3.5 Flash (batch)Google
79.0-7
GLM 5.3 (batch)Zhipu AI
78.6-8
GLM 5.2Zhipu AI
78.6-8
GLM 5.3 FlashZhipu AI
78.0-7
GLM 5.1Zhipu AI
78.0+11
GLM 5 TurboZhipu AI
78.0+43
GLM 5Zhipu AI
78.0-9
GLM 5.3 FlashXZhipu AI
77.9-9
GLM 5.3 Flash (batch)Zhipu AI
77.9-9
Qwen3.5-122B-A10BAlibaba
77.7-8
Gemma 2 27BGoogle
77.4-8
Qwen3.5-27BAlibaba
77.0-6
GPT-5 Mini (batch)OpenAI
76.8-6
GPT-5 Nano (batch)OpenAI
76.8-6
Gemini 3.5 Flash LiteGoogle
76.5-6
Gemini 3.5 Flash Lite (batch)Google
76.5-6
MiMo-V2.5-ProXiaomi
76.2-7
Qwen3.5-35B-A3BAlibaba
76.0-7
GLM 5.2 (free)Zhipu AI
75.7-7
Kimi K2.6Moonshot AI
75.7-7
Qwen3.8 Max (0902)Alibaba
75.6+120
Qwen3.7 PlusAlibaba
75.6-8
o3 MiniOpenAI
75.3-7
Claude Opus 4.1 (batch)Anthropic
75.2-7
GLM 4.7Zhipu AI
75.1+13
GLM 4.6Zhipu AI
75.1+27
GLM 4.5Zhipu AI
75.1-9
Qwen3.7 MaxAlibaba
74.8+176
Qwen3.5 Plus 2026-04-20Alibaba
74.8+187
Qwen3.6 Max PreviewAlibaba
74.8-11
Hy3Tencent
74.4-11
Claude Opus 4.1Anthropic
74.4-11
o1OpenAI
74.4-10
o3 Mini (batch)OpenAI
74.3-10
Qwen3.6 PlusAlibaba
74.1-10
MiniMax M3MiniMax
73.9-10
Claude Sonnet 4Anthropic
73.9-15
R1DeepSeek
73.8-11
o1-proOpenAI
73.6-10
Qwen3.8 27BAlibaba
73.4-10
MiMo-V2.5Xiaomi
73.0-10
Gemma 4 26B A4B Google
73.0-10
Gemma 4 26B A4B (free)Google
73.0-10
Qwen3.8 27B (free)Alibaba
72.9-10
Inklingthinkingmachines
72.9-10
Inkling (free)thinkingmachines
72.9-10
GLM 5V TurboZhipu AI
72.3-8
o4 Mini HighOpenAI
72.1-8
GPT-4o (batch)OpenAI
72.0-8
DeepSeek V3.1DeepSeek
71.8+7
DeepSeek V3 0324DeepSeek
71.8-10
Mistral Medium 3.5Mistral AI
71.6-10
MiniMax M2.7MiniMax
71.6+6
MiniMax M2.5MiniMax
71.6-11
GPT-4o (2024-11-20)OpenAI
71.2+60
GPT-4o (2024-08-06)OpenAI
71.2-12
GPT-4oOpenAI
71.2-12
GPT-4o (2024-05-13)OpenAI
71.2-12
MiniMax M2MiniMax
71.0-12
MiniMax M1MiniMax
70.8-12
GLM 4.5 AirZhipu AI
70.7-12
Llama 4 MaverickMeta
70.7-12
Claude Haiku 4.5Anthropic
69.9-9
Claude Haiku 4.5 (batch)Anthropic
69.9-9
DeepSeek V3DeepSeek
69.5-8
Qwen3 VL 235B A22B InstructAlibaba
69.3-8
DeepSeek V3.1 TerminusDeepSeek
69.3-8
GPT-4o-miniOpenAI
69.3-8
Inkling Smallthinkingmachines
68.7-8
Inkling Small (free)thinkingmachines
68.7-8
Qwen3.5-FlashAlibaba
68.6-8
Hy3 previewTencent
68.4-8
Qwen3.5 Plus 2026-02-15Alibaba
68.2+159
Qwen3 Max ThinkingAlibaba
68.2-9
MiniMax M2-herMiniMax
68.1-9
GPT-4.1OpenAI
67.7-9
GPT-4.1 (batch)OpenAI
67.7-9
Qwen3 MaxAlibaba
67.4-8
Mistral Large 3 2512 (batch)Mistral AI
67.0-8
Llama 3.3 70B InstructMeta
66.8-8
Qwen3 Next 80B A3B ThinkingAlibaba
66.7+6
Qwen3 Next 80B A3B InstructAlibaba
66.7-9
GPT-4 TurboOpenAI
66.7-9
Qwen3.5-9BAlibaba
66.5-9
Step 3.5 FlashStepFun
66.1-9
Mistral Large 2407Mistral AI
65.9+23
Mistral LargeMistral AI
65.9-10
Composer 2Cursor
65.7-10
Composer 2 FastCursor
65.7-10
GLM 4.6VZhipu AI
65.6-10
Qwen3 235B A22B Thinking 2507Alibaba
65.3-10
Llama 3.1 70B InstructMeta
65.3-10
GPT-4OpenAI
64.9-10
Qwen3 235B A22B Instruct 2507Alibaba
64.7-10
Qwen3 30B A3B Thinking 2507Alibaba
64.1-9
Qwen3 30B A3B Instruct 2507Alibaba
64.1+170
Qwen3 30B A3BAlibaba
64.1-12
o3 Mini HighOpenAI
63.9-10
GLM 4.7 FlashZhipu AI
63.5-10
Mixtral 8x22B InstructMistral AI
63.4-10
Trinity Large Thinkingarcee-ai
63.1-10
GPT-4o-mini (batch)OpenAI
62.5-10
GLM 4.5VZhipu AI
62.3-10
GPT-5 NanoOpenAI
61.9-10
gpt-oss-120bOpenAI
61.6-10
GPT-4 Turbo (batch)OpenAI
61.5-9
Mercury 2.5Inception
61.2-9
Qwen3 8BAlibaba
61.0-9
Mercury 2Inception
60.9-9
Nova 2 LiteAmazon
60.4-9
Llama 4 ScoutMeta
60.2-9
Phi 4Microsoft
60.2-9
GPT-4.1 Mini (batch)OpenAI
58.9-9
GPT-4.1 Nano (batch)OpenAI
58.9-9
gpt-oss-20bOpenAI
57.4-9
GPT-4o-mini (2024-07-18)OpenAI
56.5-10
GPT-4.1 MiniOpenAI
56.2-10
Qwen3 235B A22BAlibaba
54.0-9
Granite 4.2 8BIBM
53.8-9
GPT-5 MiniOpenAI
53.8-9
Claude 3 HaikuAnthropic
51.3-8
Command ACohere
50.8-8
Command R (08-2024)Cohere
48.7-7
Command R+ (08-2024)Cohere
48.7-8
Kimi K2.5Moonshot AI
47.5-8
Llama 3.1 8B InstructMeta
44.5-8
GPT-4.1 NanoOpenAI
42.1-8
R1 Distill Llama 70BDeepSeek
40.7-8
Ternary Bonsai 2 27Bprism-ml
40.0-19
Paretounbiased
40.0-19
DeepSeek Pro Latest~deepseek
40.0-19
DeepSeek Flash Latest~deepseek
40.0-19
Schematron V2 Turboinference-net
40.0-19
Schematron V2 Smallinference-net
40.0-19
GPT Astra Latest~openai
40.0-19
GPT Sol Latest~openai
40.0-19
GPT Terra Latest~openai
40.0-19
GPT Luna Latest~openai
40.0-19
Fugu Ultra v2sakana
40.0-19
Fugu Maxsakana
40.0-19
Ling 3.0 Flash VLinclusionai
40.0-19
Ling 3.0 Flash VL (free)inclusionai
40.0-19
DeepSeek V4.1 FlashDeepSeek
40.0-19
Ling 3.0 Flash Sante (free)inclusionai
40.0-20
Muse Spark 1.3 Contributormeta
40.0-19
Hy4 previewTencent
40.0-18
Ling 3.0 Flash Fininclusionai
40.0-18
Ling 3.0 Flash Fin (free)inclusionai
40.0-18
GLM Flash Latest~z-ai
40.0-18
Qwen3.8 FlashAlibaba
40.0-18
Muse Spark 1.2 Contributormeta
40.0-18
DeepSeek V4 Flash Vision ExpDeepSeek
40.0-18
Hy-MT2-1.8BTencent
40.0-17
Hy-MT2-30B-A3BTencent
40.0-17
GLM Latest~z-ai
40.0-17
Hy-MT2-7BTencent
40.0-17
Dots3-Note Preview (free)dots-studio
40.0-17
Gemini 3.7 FlashGoogle
40.0-17
Gemini 3.7 Flash (batch)Google
40.0-17
Seed 2.1 TurboByteDance
40.0-17
Qwen3.8 2.4T A95BAlibaba
40.0-17
Seed-2.0-CodeByteDance
40.0-16
DeepSeek V4 Pro 0813DeepSeek
40.0-16
LFM2.5-2.6B (free)Liquid AI
40.0-15
Nemotron 3.5 LightningNVIDIA
40.0-15
Nemotron 3.5 Lightning (free)NVIDIA
40.0-15
Sakana Namazusakana
40.0-15
Solar Pro 4Upstage
40.0-15
Muse Glimmer 30Bmeta
40.0-15
DeepSeek V4 Flash Latest~deepseek
40.0-13
DeepSeek V4 Flash 0731DeepSeek
40.0-13
Qwen3.7 FlashAlibaba
40.0-12
Claude Opus 5 (batch)Anthropic
40.0-11
Ling 3.0 Flashinclusionai
40.0-11
Laguna S 2.1poolside
40.0-11
Laguna S 2.1 (free)poolside
40.0-11
Gemini 3.6 FlashGoogle
40.0-11
Gemini 3.6 Flash (batch)Google
40.0-11
LongCat 2.0Meituan
40.0-11
Kimi K3Moonshot AI
40.0-11
Kimi K3 (batch)Moonshot AI
40.0-11
KAT-Coder-Pro V2.5Kuaishou
40.0-11
Grok Latest~x-ai
40.0-11
7

Leaderboard Snapshot

Top 10 AI models this week by composite score.

#ModelScore7d Change
1Claude Fable 5.1Anthropic95.90
2Claude Fable 5.1 (batch)Anthropic95.90
3Claude Fable 5Anthropic95.90
4Claude Fable 5 (batch)Anthropic95.90
5Claude Opus 5.5Anthropic95.1+426
6Claude Opus 5Anthropic95.1+272
7Claude Opus 4.8Anthropic95.10
8Claude Opus 4.7Anthropic95.1-3
9Claude Opus 4.7 (batch)Anthropic95.1-3
10Claude Opus 4.8 (batch)Anthropic94.6-2
8

This Week's Numbers

Total Ranked

300

Avg Score

68.2

Top 10 Moved

5/10

Most Active Provider

OpenAI

9

Read More Analysis

Degradation Tracker
Stability Report
Significance Tracker
Trending Models
Frequently Asked Questions

The weekly report summarizes seven days of AI model ranking changes. It highlights the biggest gainers and losers by rank position, new models entering the leaderboard, models in a fragile watch-list state, and a snapshot of the current top 10. All data comes from hourly-updated composite scores across 290+ models.

Gainers and losers are ranked by their 7-day rank change, which measures how many positions a model moved up or down on the leaderboard over the past week. The models with the largest positive changes are the top gainers, and those with the largest negative changes are the top losers.

The watch list contains models currently in a "fragile" state, meaning their rankings are unstable and could degrade further. Monitoring these models is important if you rely on them in production, as they may experience significant quality or performance drops in the near term.

Weekly AI Performance Report - This Week in AI Models | LM Market Cap