Qwen2.5-7B-Instruct
Qwen2.5-7B-Instruct is the instruction-tuned 7B language model released with Qwen2.5 on September 19, 2024. Its advertised 131,072-token context requires YaRN for inputs above the default 32,768-token configuration.
Benchmark scores
Access
Open-weight release under the exact repository’s Apache 2.0 licence. Advertised full context is 131,072 tokens, but the card states a 32,768-token default configuration and requires YaRN beyond that; generation is specified as 8,192 tokens. Hosted API/GA status, exact inference budgets and runtime performance are unverified. Qwen-Plus and Qwen-Turbo APIs are separate identities.