AI Model Leaderboard

Ranked on real OrcaRouter production traffic and community Battle Mode votes — not vendor-reported benchmarks.

Measured on real traffic
100.0%
Top success rate (7d)
Google: Nano Banana Pro (Gemini 3 Pro Image Preview)
414 ms
Fastest p50 latency
Google: Gemini 2.5 Flash Lite
80.1%
Top community win rate
Qwen3.7 Max (2026-05-20)

Overall — community win rate

Blind Battle Mode votes, ranked by a Bradley–Terry (Elo) model. Dots show the rating; bars show the 95% confidence interval. Models within a band are statistically tied.

Style control
Coming soon — needs more battle data

⚔️ Blind Battle

Two anonymous models answer the same prompt. Read both, vote the winner, then see who wrote what — your vote feeds the ranking above.

RankModelArena RatingMethodVotesW·L·T
#1Qwen3.7 Max (2026-05-20)1892±140
blind · BT
168129·32·7
#2DeepSeek: DeepSeek V4 Pro1673±74
blind · BT
252163·77·12
OpenAI: GPT-5.4 Pro1669±79
blind · BT
281158·111·12
OpenAI: GPT-5.51667±79
blind · BT
230109·109·12
Qwen: Qwen3.5-35B-A3B1666±88
blind · BT
310149·149·12
Qwen3.7 Max1665±125
blind · BT
22484·130·10
Anthropic: Claude Opus 4.71664±102
blind · BT
247117·117·13
Qwen3.6 35B A3B Uncensored (Aggressive)1664±90
blind · BT
260124·124·12
Anthropic: Claude Opus 4.81663±95
blind · BT
245116·116·13
Google: Gemma 4 26B A4B1662±90
blind · BT
242114·116·12
Qwen: Qwen3.6 Plus1661±124
blind · BT
278108·158·12
#12OpenAI: GPT-5.5 Pro1445±67
blind · BT
29495·187·12
MiniMax: MiniMax M31445±58
blind · BT
266128·126·12
MoonshotAI: Kimi K2.7 Code1445±69
blind · BT
224119·93·12
Google: Gemma 4 31B1445±58
blind · BT
273112·149·12
Qwen: Qwen3.5-27B1442±64
blind · BT
213100·101·12
Z.ai: GLM 5.21440±67
blind · BT
235141·82·12
#18Google: Nano Banana 2 (Gemini 3.1 Flash Image Preview)1228±104
blind · BT
281151·118·12
OpenAI: GPT-5.4 Nano1227±98
blind · BT
308196·100·12
Google: Gemini 3.1 Flash Lite Preview1226±96
blind · BT
285137·136·12
MiniMax M2.7 highspeed1226±103
blind · BT
246117·117·12
MiniMax: MiniMax M2.71225±66
blind · BT
22681·133·12
OpenAI: GPT-5.41224±74
blind · BT
245117·116·12
Qwen: Qwen3.7 Plus1224±89
blind · BT
248118·118·12
Z.ai: GLM 5.11222±77
blind · BT
221105·104·12
Kling: Kling 3.0 Turbo1222±88
blind · BT
256121·123·12
Qwen: Qwen3.6 Flash1218±71
blind · BT
18871·105·12
#28Gemini 3.5 Flash1004±119
blind · BT
28596·177·12
DeepSeek: DeepSeek V4 Flash1004±124
blind · BT
13061·62·7
OpenAI: GPT-5.4 Mini1003±120
blind · BT
20675·122·9

Head-to-head

Pairwise Battle Mode win rates — how often each model beats each rival head-to-head.

Category leaders

The strongest model in each capability — coding, math, reasoning and more — with a per-axis profile against the field median.

Benchmarks — intelligence vs price

Independent intelligence scores plotted against input price. The dashed line is the price/intelligence Pareto frontier.

Reliability & cost

Measured on OrcaRouter production traffic over the last 7 days.

Model
Success % (7d)
p50
p99
Err %
$/1M (in→out)
tok/s
Google: Nano Banana Pro (Gemini 3 Pro Image Preview)100.0%4.29 s4.29 s
0.0%
— → —
OpenAI: GPT-3.5 Turbo100.0%1.57 s4.34 s
0.0%
$0.50 → $1.50149
Qwen: Qwen3 VL 235B A22B Thinking100.0%4.00 s10.00 s
0.0%
$0.40 → $4.0036
openai/gpt-5.2-2025-12-11100.0%1.13 s5.64 s
0.0%
$1.75 → $14.0099
OpenAI: GPT-5.2-Codex100.0%607 ms2.20 s
0.0%
$1.75 → $14.0079
openai/gpt-3.5-turbo-1106100.0%1.82 s10.00 s
0.0%
$1.00 → $2.00135
openai/gpt-4o-mini-search-preview-2025-03-11100.0%3.50 s3.75 s
0.0%
$0.15 → $0.6040
Anthropic: Claude Haiku 4.5100.0%6.58 s10.00 s
0.0%
$1.00 → $5.00126
openai/gpt-4-0613100.0%1.00 s1.11 s
0.0%
$30.00 → $60.00
openai/gpt-4-turbo-2024-04-09100.0%4.06 s10.00 s
0.0%
$10.00 → $30.0031
openai/gpt-5-2025-08-07100.0%10.00 s10.00 s
0.0%
$1.25 → $10.00323
openai/gpt-5.1-2025-11-13100.0%1.36 s10.00 s
0.0%
$1.25 → $10.00105
openai/gpt-5-nano-2025-08-07100.0%5.00 s10.00 s
0.0%
$0.05 → $0.40565
OpenAI: GPT-4 Turbo100.0%5.50 s10.00 s
0.0%
$10.00 → $30.0033
Qwen: Qwen3.5-27B100.0%6.67 s10.00 s
0.0%
$0.09 → $0.6956
Google: Gemini 3 Flash Preview100.0%4.81 s10.00 s
0.0%
$0.50 → $3.00995
Google: Gemini 3.1 Flash Lite Preview100.0%836 ms5.62 s
0.0%
$0.25 → $1.50222
Qwen: Qwen3 VL 8B Instruct100.0%3.64 s10.00 s
0.0%
$0.18 → $0.7053
OpenAI: GPT-4.1 Mini100.0%1.33 s7.17 s
0.0%
$0.40 → $1.6063
OpenAI: GPT-5.4 Mini100.0%1.56 s10.00 s
0.0%
$0.75 → $4.50220
codex-auto-review100.0%2.93 s10.00 s
0.0%
$2.50 → $15.0026
Google: Gemini 2.5 Pro100.0%7.78 s10.00 s
0.0%
$2.50 → $15.00411
OpenAI: GPT-5.4100.0%6.73 s10.00 s
0.0%
$5.00 → $22.5072
openai/gpt-5.4-2026-03-05100.0%1.22 s2.62 s
0.0%
$5.00 → $22.50171
google/gemini-flash-latest100.0%2.33 s10.00 s
0.0%
$0.50 → $3.00534
Showing 1 to 25 of 143

Volume — production traffic

Which models actually carry OrcaRouter production traffic, ranked by token throughput over the selected window.

Volume ranking is warming up

We have seen — models so far this period. The volume ranking unlocks once enough traffic has accumulated to rank models fairly.

Models seen

In the meantime, the Overall and Reliability boards above are already live.

Validation — agreement with external rankings

How closely the community ranking tracks independent external rankings, measured by Spearman’s ρ and Kendall’s τ rank correlation.

Buzz — popularity & sentiment

What the community is talking about — mention volume, 14-day momentum and sentiment. Context only, never a ranking signal.