Vibe Bench live
Reading the market
Reading verified source signals
The daily market board
Daily AI benchmark changes
Track the models gaining and losing ground each day. The Vibe Coding Index, or VCI, combines current intelligence, coding, and agentic evidence on one scale.
Latest daily close
Jul 16, 2026
8 PM CT snapshot
Refreshes at the top of every hour
Live benchmark data
122 models tracked
Kimi K3 leads at 67.3 0 latest winners
0 latest losers
30 days of source history Hourly source refresh
Current leader
Kimi K367.3 VCI
Latest winners
0
Models moving up
Latest losers
0
Models moving down
Models tracked
122
122 unchanged today
Showing verified movement since the previous hourly refresh
Open every recorded snapshot →
Biggest gains
Latest winners
No models moved in this direction during the latest comparison.
Biggest declines
Latest losers
No models moved in this direction during the latest comparison.
Real source history
30-day usage momentum
Official daily token traffic from OpenRouter, backfilled for up to 30 completed days. This measures model usage and market attention—it is displayed beside VCI movement but never changes a model’s quality score.
History loaded
30
completed days
Source window
Jun 17 → Jul 16
UTC daily closes
Matched models
39
on latest source day
Gaining attention
Usage risers
MiMo-V2.5#2 usage rank · +2 places
+7.091share pts
Nemotron 3 Ultra 550B A55B#6 usage rank · +12 places
+3.310share pts
GLM-5.2#4 usage rank · +7 places
+3.152share pts
Claude Sonnet 5#11 usage rank
+1.682share pts
GPT-5.6 Sol#22 usage rank
+0.804share pts
Claude Fable 5#23 usage rank
+0.637share pts
Losing share
Usage fallers
Hy3Outside latest top 50
-7.452share pts
GLM-5.1#46 usage rank · -39 places
-4.284share pts
MiniMax-M3#5 usage rank · -3 places
-2.848share pts
Claude Opus 4.7#8 usage rank · -3 places
-1.919share pts
GPT-5.5#15 usage rank · -3 places
-1.880share pts
Claude Sonnet 4.6#10 usage rank · -1 places
-1.858share pts
OpenRouter reports the top 50 public models per completed UTC day plus an aggregated “other” row. Shares use the full reported daily total, model variants are collapsed to a stable base model before ranking. Source window. 2026-06-17 through 2026-07-16.
The full daily board
Every tracked model and its movement
The trend line uses one closing score per Central Time calendar day and covers up to seven days.
| Rank | Model | Current VCI | Latest change | Rank change | Seven day change | Seven day trend |
|---|---|---|---|---|---|---|
| 1 |
Kimi K3Moonshot |
67.3 | 0.0 | 0 | — | Baseline |
| 2 |
Claude Fable 5Anthropic |
67.0 | 0.0 | 0 | — | Baseline |
| 3 |
GPT-5.6 SolOpenAI |
66.5 | 0.0 | 0 | — | Baseline |
| 4 |
GPT-5.6 TerraOpenAI |
62.1 | 0.0 | 0 | — | Baseline |
| 5 |
Claude Opus 4.8Anthropic |
62.0 | 0.0 | 0 | — | Baseline |
| 6 | 
Grok 4.5xAI |
60.5 | 0.0 | 0 | — | Baseline |
| 7 |
Claude Opus 4.7Anthropic |
60.3 | 0.0 | 0 | — | Baseline |
| 8 |
Claude Sonnet 5Anthropic |
60.1 | 0.0 | 0 | — | Baseline |
| 9 |
GPT-5.5OpenAI |
59.9 | 0.0 | 0 | — | Baseline |
| 10 |
GLM-5.2Zhipu |
58.7 | 0.0 | 0 | — | Baseline |
| 11 |
GPT-5.6 LunaOpenAI |
58.3 | 0.0 | 0 | — | Baseline |
| 12 | 
Muse Spark 1.1Meta |
56.2 | 0.0 | 0 | — | Baseline |
| 13 |
GPT-5.4OpenAI |
55.8 | 0.0 | 0 | — | Baseline |
| 14 | 
Gemini 3.5 FlashGoogle |
54.7 | 0.0 | 0 | — | Baseline |
| 15 |
Claude Sonnet 4.6Anthropic |
53.9 | 0.0 | 0 | — | Baseline |
| 16 | 
Qwen3.7 MaxAlibaba |
50.8 | 0.0 | 0 | — | Baseline |
| 17 |
MiniMax-M3MiniMax |
49.5 | 0.0 | 0 | — | Baseline |
| 18 | 
DeepSeek V4 ProDeepSeek |
49.3 | 0.0 | 0 | — | Baseline |
| 19 |
Kimi K2.6Moonshot |
49.1 | 0.0 | 0 | — | Baseline |
| 20 |
Kimi K2.7 CodeMoonshot |
47.1 | 0.0 | 0 | — | Baseline |
| 21 |
Gemini 3.1 Pro PreviewGoogle |
46.9 | 0.0 | 0 | — | Baseline |
| 22 |
MiMo-V2.5-ProXiaomi |
46.9 | 0.0 | 0 | — | Baseline |
| 23 |
GLM-5.1Zhipu |
46.7 | 0.0 | 0 | — | Baseline |
| 24 | 
Nex-N2-ProNex Agi |
45.6 | 0.0 | 0 | — | Baseline |
| 25 |
Hy3Tencent |
45.4 | 0.0 | 0 | — | Baseline |
| 26 |
DeepSeek V4 FlashDeepSeek |
44.2 | 0.0 | 0 | — | Baseline |
| 27 |
GPT-5.4 miniOpenAI |
43.9 | 0.0 | 0 | — | Baseline |
| 28 |
Qwen3.6 PlusAlibaba |
43.9 | 0.0 | 0 | — | Baseline |
| 29 |
GPT-5.4 nanoOpenAI |
42.5 | 0.0 | 0 | — | Baseline |
| 30 |
MiMo-V2.5Xiaomi |
42.0 | 0.0 | 0 | — | Baseline |
| 31 |
Qwen3.6 27BAlibaba |
41.0 | 0.0 | 0 | — | Baseline |
| 32 |
Grok Build 0.1 0616xAI |
40.9 | 0.0 | 0 | — | Baseline |
| 33 |
MiniMax-M2.7MiniMax |
40.8 | 0.0 | 0 | — | Baseline |
| 34 |
Qwen3.7 PlusAlibaba |
40.2 | 0.0 | 0 | — | Baseline |
| 35 |
Claude 4.5 SonnetAnthropic |
39.3 | 0.0 | 0 | — | Baseline |
| 36 |
Nemotron 3 Ultra 550B A55BNVIDIA |
39.3 | 0.0 | 0 | — | Baseline |
| 37 | 
KAT Coder Pro V2Kwaipilot |
38.9 | 0.0 | 0 | — | Baseline |
| 38 |
GLM-4.7Zhipu |
38.8 | 0.0 | 0 | — | Baseline |
| 39 |
Kimi K2.5Moonshot |
38.1 | 0.0 | 0 | — | Baseline |
| 40 |
GPT-5.1OpenAI |
38.0 | 0.0 | 0 | — | Baseline |
| 41 |
Qwen3.5 397B A17BAlibaba |
36.7 | 0.0 | 0 | — | Baseline |
| 42 |
Grok 4.3xAI |
36.4 | 0.0 | 0 | — | Baseline |
| 43 |
GPT-5OpenAI |
35.9 | 0.0 | 0 | — | Baseline |
| 44 |
Qwen3.5 122B A10BAlibaba |
35.3 | 0.0 | 0 | — | Baseline |
| 45 |
DeepSeek V3.2DeepSeek |
34.0 | 0.0 | 0 | — | Baseline |
| 46 |
GLM-4.6Zhipu |
33.3 | 0.0 | 0 | — | Baseline |
| 47 |
Qwen3.6 35B A3BAlibaba |
32.7 | 0.0 | 0 | — | Baseline |
| 48 | 
Mistral Medium 3.5Mistral |
32.3 | 0.0 | 0 | — | Baseline |
| 49 |
DeepSeek V3.1 TerminusDeepSeek |
32.0 | 0.0 | 0 | — | Baseline |
| 50 | 
Ring-2.6-1TInclusionai |
32.0 | 0.0 | 0 | — | Baseline |
| 51 |
Gemma 4 31BGoogle |
31.9 | 0.0 | 0 | — | Baseline |
| 52 |
Claude 4.5 HaikuAnthropic |
31.4 | 0.0 | 0 | — | Baseline |
| 53 | 
Step 3.7 FlashStepfun |
31.4 | 0.0 | 0 | — | Baseline |
| 54 |
MiMo-V2-FlashXiaomi |
30.5 | 0.0 | 0 | — | Baseline |
| 55 |
Gemma 4 26B A4BGoogle |
28.6 | 0.0 | 0 | — | Baseline |
| 56 |
Claude 4 SonnetAnthropic |
28.5 | 0.0 | 0 | — | Baseline |
| 57 |
NVIDIA Nemotron 3 Super 120B A12BNVIDIA |
25.1 | 0.0 | 0 | — | Baseline |
| 58 |
Qwen3 Coder NextAlibaba |
23.6 | 0.0 | 0 | — | Baseline |
| 59 |
gpt-oss-120bOpenAI |
23.1 | 0.0 | 0 | — | Baseline |
| 60 |
Gemini 3.1 Flash-LiteGoogle |
22.8 | 0.0 | 0 | — | Baseline |
| 61 |
Gemini 2.5 ProGoogle |
22.0 | 0.0 | 0 | — | Baseline |
| 62 | 
North Mini CodeCohere |
21.5 | 0.0 | 0 | — | Baseline |
| 63 |
Devstral 2Mistral |
21.2 | 0.0 | 0 | — | Baseline |
| 64 |
Devstral Small 2Mistral |
20.3 | 0.0 | 0 | — | Baseline |
| 65 |
Command A+Cohere |
20.2 | 0.0 | 0 | — | Baseline |
| 66 |
Qwen3.5 9BAlibaba |
19.8 | 0.0 | 0 | — | Baseline |
| 67 |
GPT-5 miniOpenAI |
18.9 | 0.0 | 0 | — | Baseline |
| 68 |
Mistral Small 4Mistral |
17.5 | 0.0 | 0 | — | Baseline |
| 69 |
Mistral Small 3.1Mistral |
16.6 | 0.0 | 0 | — | Baseline |
| 70 |
Mistral Large 3Mistral |
16.0 | 0.0 | 0 | — | Baseline |
| 71 |
DeepSeek R1DeepSeek |
15.9 | 0.0 | 0 | — | Baseline |
| 72 |
Qwen3 235B A22B 2507Alibaba |
15.2 | 0.0 | 0 | — | Baseline |
| 73 | 
Nova 2.0 LiteAmazon |
15.1 | 0.0 | 0 | — | Baseline |
| 74 |
Ling 2.6 FlashInclusionai |
15.0 | 0.0 | 0 | — | Baseline |
| 75 |
Mistral Medium 3.1Mistral |
14.3 | 0.0 | 0 | — | Baseline |
| 76 |
DeepSeek V3DeepSeek |
13.8 | 0.0 | 0 | — | Baseline |
| 77 |
K2 Think V2Moonshot |
13.5 | 0.0 | 0 | — | Baseline |
| 78 |
gpt-oss-20bOpenAI |
13.4 | 0.0 | 0 | — | Baseline |
| 79 |
DeepSeek V3 0324DeepSeek |
13.1 | 0.0 | 0 | — | Baseline |
| 80 |
GPT-4.1 miniOpenAI |
12.6 | 0.0 | 0 | — | Baseline |
| 81 |
Qwen3 Next 80B A3BAlibaba |
11.9 | 0.0 | 0 | — | Baseline |
| 82 |
o3-miniOpenAI |
11.1 | 0.0 | 0 | — | Baseline |
| 83 | 
Solar Pro 3Upstage |
11.1 | 0.0 | 0 | — | Baseline |
| 84 |
Llama 4 MaverickMeta Llama |
10.7 | 0.0 | 0 | — | Baseline |
| 85 |
Claude 3.5 HaikuAnthropic |
10.3 | 0.0 | 0 | — | Baseline |
| 86 |
NVIDIA Nemotron 3 Nano 30B A3BNVIDIA |
10.0 | 0.0 | 0 | — | Baseline |
| 87 |
Qwen3 32BAlibaba |
9.8 | 0.0 | 0 | — | Baseline |
| 88 |
Ministral 3 14BMistral |
9.5 | 0.0 | 0 | — | Baseline |
| 89 |
Qwen3 30B A3B 2507Alibaba |
9.0 | 0.0 | 0 | — | Baseline |
| 90 |
Qwen3 14BAlibaba |
8.9 | 0.0 | 0 | — | Baseline |
| 91 |
Llama 3.3 Instruct 70BMeta Llama |
7.3 | 0.0 | 0 | — | Baseline |
| 92 |
GPT-4.1 nanoOpenAI |
7.3 | 0.0 | 0 | — | Baseline |
| 93 |
Ministral 3 8BMistral |
6.6 | 0.0 | 0 | — | Baseline |
| 94 |
Qwen3 8BAlibaba |
6.2 | 0.0 | 0 | — | Baseline |
| 95 |
Gemma 3 27B InstructGoogle |
6.1 | 0.0 | 0 | — | Baseline |
| 96 |
Llama 4 ScoutMeta Llama |
6.1 | 0.0 | 0 | — | Baseline |
| 97 |
Llama 3.1 Instruct 8BMeta Llama |
4.1 | 0.0 | 0 | — | Baseline |
| 98 |
Ministral 3 3BMistral |
4.1 | 0.0 | 0 | — | Baseline |
| 99 |
Gemma 3 12B InstructGoogle |
3.8 | 0.0 | 0 | — | Baseline |
| 100 | 
Phi-4 Mini InstructMicrosoft |
3.0 | 0.0 | 0 | — | Baseline |
| — |
Gemini 2.5 Pro PreviewGoogle |
— | — | — | — | Baseline |
| — |
o1OpenAI |
— | — | — | — | Baseline |
| — |
Claude 3.5 SonnetAnthropic |
— | — | — | — | Baseline |
| — |
Claude 3.5 Sonnet (June '24)Anthropic |
— | — | — | — | Baseline |
| — |
GPT-4oOpenAI |
— | — | — | — | Baseline |
| — |
Gemini 1.5 ProGoogle |
— | — | — | — | Baseline |
| — |
GPT-4 TurboOpenAI |
— | — | — | — | Baseline |
| — |
Claude 3 OpusAnthropic |
— | — | — | — | Baseline |
| — |
Claude 2.1Anthropic |
— | — | — | — | Baseline |
| — |
GPT-4OpenAI |
— | — | — | — | Baseline |
| — |
Claude 2.0Anthropic |
— | — | — | — | Baseline |
| — |
Gemini 3.5 ProGoogle |
— | — | — | — | Baseline |
| — |
Claude Mythos 5Anthropic |
— | — | — | — | Baseline |
| — |
Claude Opus 5Anthropic |
— | — | — | — | Baseline |
| — |
GLM-5.3Zhipu |
— | — | — | — | Baseline |
| — |
DeepSeek R2DeepSeek |
— | — | — | — | Baseline |
| — |
GPT-4o miniOpenAI |
— | — | — | — | Baseline |
| — |
GPT-3.5 TurboOpenAI |
— | — | — | — | Baseline |
| — |
Granite 4.1 8BIbm Granite |
— | — | — | — | Baseline |
| — |
Claude InstantAnthropic |
— | — | — | — | Baseline |
| — |
Gemma 3n E4B InstructGoogle |
— | — | — | — | Baseline |
| — |
Gemma 3 4B InstructGoogle |
— | — | — | — | Baseline |
One close per day
The latest successful refresh becomes the daily close. Every earlier refresh still remains available in benchmark history.
No invented VCI movement
A quality winner or loser appears only when two recorded VCI closes can be compared. Backdated usage remains separately labeled.
Evidence stays visible
Every model links to its scores, source coverage, confidence, and complete recorded history.