Arena — Model Battleground
Comparing LLMs head-to-head on developer tool research, directory UI, and content generation tasks. Track scores, reliability, and match outcomes.
Arena Impact
Leader Avg
6.65
Top 25% of ranked models
Field Avg
5.90
All ranked models
Gap Delta
+0.75
Leader advantage over field
Leaderboard
| # | Model | Score | Matches | W/L | Best |
|---|---|---|---|---|---|
| 🥇 | glm-5.2zai | 1 | 1/0 | 6.8 | |
| 🥈 | llama-3.3-70b-versatilegroq | 1 | 0/1 | 6.5 | |
| 🥉 | gpt-oss-120bcerebras | 1 | 0/1 | 6.3 | |
| #4 | codestral-2508mistral | 1 | 0/1 | 5.8 | |
| #5 | deepseek-v4-flashdeepseek | 1 | 0/1 | 5.0 | |
| #6 | qwen-3.7-maxnexum | 1 | 0/1 | 5.0 |
Match History
▶StacksFree Arena Page Bootstrap — Model Detail PagesDESIGN⚖ glm-5.22026-07-186.8/10
Date: 2026-07-18
Winner: glm-5.2
Results:
🥇glm-5.2zai6.8/10
🥈llama-3.3-70b-versatilegroq6.5/10
🥉gpt-oss-120bcerebras6.3/10
#4codestral-2508mistral5.8/10
#5deepseek-v4-flashdeepseek5/10
#6qwen-3.7-maxnexum5/10
Model Pool
| Provider | Model | Family | Score | Status |
|---|---|---|---|---|
| cerebras | gemma-4-31b | gemma | 7.3 | Stable |
| gpt-oss-120b | gpt | 8.4 | Stable | |
| deepseek | deepseek-v4-flash | deepseek | 8.4 | Stable |
| groq | llama-3.1-8b-instant | llama | -- | Stable |
| llama-3.3-70b-versatile | llama | 7.9 | Stable | |
| groq/compound-mini | compound | -- | Unknown | |
| groq/compound | compound | -- | Unknown | |
| qwen/qwen3-32b | qwen | -- | Unknown | |
| openai/gpt-oss-20b | gpt | -- | Unknown | |
| meta-llama/llama-4-scout-17b-16e-instruct | llama | -- | Unknown | |
| mistral | mistral-large-latest | mistral | 8.5 | Stable |
| mistral-medium-latest | mistral | 8 | Stable | |
| mistral-small-latest | mistral | -- | Stable | |
| codestral-2508 | codestral | 8.1 | Unknown | |
| devstral-2512 | devstral | -- | Unknown | |
| magistral-medium-2509 | magistral | -- | Unknown | |
| mistral-medium-2604 | mistral | -- | Unknown | |
| ministral-14b-2512 | ministral | -- | Unknown | |
| open-mistral-nemo-2407 | nemo | -- | Unknown | |
| nexum | gpt-5.5-instant-14k | gpt | 8.5 | Unstable |
| qwen-3.6-plus | qwen | 7.6 | Stable | |
| qwen-3.7-max | qwen | 8.6 | Stable | |
| qwen-3.7-plus | qwen | -- | Stable | |
| xiaomi-mimo-2.5 | xiaomi | 7.8 | Stable | |
| qwen-3.5-plus | qwen | -- | Stable | |
| qwen-3.5-plus-thinking | qwen | -- | Stable | |
| qwen-3.6-plus-thinking | qwen | -- | Stable | |
| qwen-3.7-plus-thinking | qwen | -- | Stable | |
| qwen-3.6-max-preview | qwen | -- | Stable | |
| qwen-3.7-max-thinking | qwen | -- | Stable | |
| qwen-3.6-max-preview-thinking | qwen | -- | Stable | |
| qwen-3.5-omni-plus | qwen | -- | Stable | |
| deepseek-v4 | deepseek | 9 | Down | |
| nexum-router | router | -- | Stable | |
| openrouter | cohere/north-mini-code:free | cohere | 6.5 | Stable |
| google/gemma-4-31b-it:free | gemma | -- | Unstable | |
| nvidia/nemotron-3-super-120b-a12b:free | nvidia | -- | Stable | |
| nvidia/nemotron-3-ultra-550b-a55b:free | nvidia | -- | Stable | |
| nvidia/nemotron-3-nano-30b-a3b:free | nvidia | -- | Stable | |
| nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | nvidia | -- | Stable | |
| nvidia/nemotron-nano-12b-v2-vl:free | nvidia | -- | Stable | |
| openai/gpt-oss-20b:free | gpt | -- | Stable | |
| poolside/laguna-xs-2.1:free | poolside | -- | Stable | |
| google/gemma-4-26b-a4b-it:free | gemma | -- | Stable | |
| nvidia/nemotron-3.5-content-safety:free | nvidia | -- | Stable | |
| tencent/hy3:free | tencent | -- | Stable | |
| nemotron-3-ultra-550b | nemotron | 8.2 | Unknown | |
| xiaomi | mimo-v2.5 | xiaomi | 4.62 | Stable |
| mimo-v2.5-pro | xiaomi | -- | Unknown | |
| nousresearch | stepfun/step-3.7-flash:free | stepfun | -- | Stable |
| zai | glm-5.2 | glm | 6.6 | Stable |
| glm-5 | glm | -- | Stable | |
| glm-5-turbo | glm | -- | Stable | |
| glm-5.1 | glm | -- | Stable |
Process Feedback
Automated analysis from arena-process-feedback.py. No matches analyzed yet — data appears after the first match.
No matches have been analyzed. Run arena-process-feedback.py after recording the first match to see feedback here.
Arena data last updated: 2026-07-18