Instruction tuning with SFT and DPO
According to the model card, this repo contains the instruction-tuned 1.5B Qwen2 model, post-trained with both supervised fine-tuning and direct preference optimization.
Open Source Model Profile · Qwen
Qwen2-1.5B-Instruct
Qwen2-1.5B-Instruct is a 1.54B-parameter instruction-tuned Qwen2 model from Qwen for text generation. According to the model card, it was post-trained with supervised fine-tuning and direct preference optimization.
- Publisher
- Qwen
- Task
- text-generation
- Model type
- qwen2
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Qwen2-1.5B-Instruct is published by Qwen as an instruction-tuned member of the Qwen2 series, which spans 0.5B to 72B sizes including a Mixture-of-Experts model. The captured configuration identifies Qwen2ForCausalLM with model type qwen2, and Safetensors metadata reports 1,543,714,304 parameters. According to the model card, Qwen2 uses a Transformer architecture with SwiGLU activation, attention QKV bias, and group query attention, plus a tokenizer adapted to multiple natural languages and code.
Recorded capabilities
Qwen2 decoder architecture
According to the model card, the series uses a Transformer decoder with SwiGLU activation, attention QKV bias, and group query attention, with an improved multilingual and code-adapted tokenizer.
Publisher-reported benchmark table
According to the model card, Qwen2-1.5B-Instruct reports MMLU 52.4, HumanEval 37.8, GSM8K 61.6, C-Eval 63.8, and IFEval prompt strict-accuracy 29.0, ahead of the listed Qwen1.5-1.8B-Chat figures.
Captured 1.54B Transformers weights
Safetensors metadata reports 1,543,714,304 parameters, and the hub lists Transformers support for text generation.
Use cases in the source record
- Conversational text-generation workflows using the instruction-tuned 1.5B model with Transformers and the documented chat-template path.
- Benchmark-referenced comparison work against Qwen1.5-1.8B-Chat on language, coding, and math tasks, using the publisher's reported results table.
Limitations and unknowns
- No context-window value was extracted from this record.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
- Competitiveness and benchmark claims are publisher model-card statements and were not independently verified by Ethen.
Source and provenance
Source: Qwen/Qwen2-1.5B-Instruct
Captured: Unknown. Processed: 2026-09-07T19:34:36.065300+00:00.
Qwen2-1.5B-Instruct Introduction Qwen2 is the new series of Qwen large language models. For Qwen2, we release a number of base language models and instruction-tuned language models ranging from 0.5 to 72 billion parameters, including a Mixture-of-Experts model. This repo contains the instruction-tuned 1.5B Qwen2 model. Compared with the state-of-the-art opensource language models, including the previous released Qwen1.5, Qwen2 has generally surpassed most opensource models and demonstrated competitiveness against proprietary models across a series of benchmarks targeting for language understanding, language generation, multilingual…
F001F002F003F004F005F006F007F008F010F011F012F014F015F016F017F018