Mixtral 8x22B Instruct
53.7%
no interval published
aiqbench · LLM benchmarks
Benchmarks
- Sources
- 8
- Entries tracked
- 22,308
- Last sync
- 17 min17 minutes ago
All sources syncing normally
- Arena · Text· 10 hours ago
- Arena · WebDev· 10 hours ago
- Arena · Vision· 10 hours ago
- Arena · Agent· 10 hours ago
- GPQA Diamond· 17 minutes ago
- SWE-bench· 17 minutes ago
- Terminal-Bench· 17 minutes ago
- ARC-AGI-2· 17 minutes ago
Filter by org:
Released:
One grid, every column. Pick axes in the Axes panel — one axis ranks like a single board, several blend into a weighted composite, all of them read as the index. Rank is assigned before any filter and never renumbers.
| Model | Org | ± CI | Swing | Add to compare | |||||
|---|---|---|---|---|---|---|---|---|---|
| 213 | Mixtral 8x22B Instruct | mistral | Apr 2024 | $2 / $6 | 53.7% | — | — |
Showing MMLU-Pro (Artificial Analysis) exactly as the source published it — its own score and confidence interval, no blending. Select a second axis to build a weighted composite.