Aya Expanse 32B
Cohere·
LLMsopen-weightcloud + localEdge / Consumer
Overview
32B multilinguales Open-Weight-Modell aus der Aya-Forschung von Cohere Labs, 23 Sprachen, Schwerpunkt auf non-English-Evaluierungen (Global-MMLU, mArenaHard). Die englische Standard-Suite reportet der Vendor nicht, multilinguale Stärke und offene Gewichte sind der Fokus. Vectara HHEM 10.9% Halluzination.
Capabilities and innovations
32B DenseMultilingual (23 Sprachen)Open weightsAya-ForschungslinieMultilingual-fokussiertes Open-Weight-ModellData Arbitrage und multilinguales Preference-Training
Benchmarks
No benchmark values are recorded for this model.
- GPQA Diamond: not reported by the vendor — Aya-Linie reportet multilinguale Evals (Global-MMLU 66.9% 5-shot, mArenaHard), nicht die englische Standard-Suite.
- Humanity’s Last Exam (no tools): benchmark did not exist at release
- SWE-bench Verified: not reported by the vendor
- SWE-bench Pro: benchmark did not exist at release
- MMLU: not reported by the vendor
- MMLU-Pro: not reported by the vendor
- Terminal-Bench 2.x: benchmark did not exist at release
- MMMU-Pro: benchmark did not exist at release
Architecture and hardware
- Parameters
- 32B total, 32B active per token (Dense)
- Estimated VRAM at Q4
- ~19 GB, Edge / Consumer class
Links
More from Cohere
Data curated by AI Model Timeline. See the methodology for admission criteria, benchmark eras and source priority.