Orca 2 13B
13B reasoning model teaching small LMs to pick the best strategy per task.
About
Orca 2 13B is a 13-billion-parameter language model released by Microsoft Research in November 2023. It builds on Meta's LLaMA 2 13B base and is fine-tuned with a cautious reasoning methodology that teaches the model to select the best reasoning strategy for each task, such as step-by-step processing, recall-then-generate, or direct answer.
Microsoft showed it outperforming LLaMA-2-Chat-13B by a wide margin and matching models 5 to 10 times larger on complex zero-shot reasoning benchmarks. The weights are publicly released under the Microsoft Research License, which limits use to non-commercial research.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 123; Trending score 62/100 from researcher assessment.