Qwen 1.5
by Alibaba Cloud (Qwen team)
Retired, open-weights text-only LLM series.
Score
Score
Our verdict
Qwen 1.5 is an open-weights, text-only LLM family from Alibaba Cloud that landed early 2024, spanning 0.5B–110B plus an MoE A2.7B. It delivers solid multilingual, code, and math performance for its generation and is easy to run locally thanks to first-class support in Transformers, vLLM, SGLang, llama.cpp, Ollama, and LM Studio, alongside official GPTQ/AWQ/GGUF quantizations. The big caveat in 2026 is lifecycle: Alibaba retired all hosted Qwen 1.5 endpoints in Aug 2025, and Qwen itself calls 1.5 the beta of Qwen2. Practically, that means there’s no vendor SLA or security updates, and only self-hosting remains. Licensing also demands care: dense models use the custom Tongyi Qianwen license (not OSI-approved), with Apache-2.0 only for the MoE A2.7B variant. If you need a managed API or modern features like longer contexts or multimodality, look to qwen-plus or newer Qwen2/2.5/3 models; otherwise, Qwen 1.5 still works for cost-free, local text generation and fine-tuning.
Overview
Score breakdown
Overall score
Scores are editorial assessments by the Compare Best AI team on a 0–10 scale.
Expert review
CBAI Editorial Team
Compare Best AI · Editorial Team
## Overview
How we tested
Days tested
7 days
Tasks evaluated
- ·Local inference via Transformers and vLLM across 7B/14B
- ·Quantized GGUF runs with llama.cpp and Ollama
- ·Instruction-tuning a 7B variant using Axolotl
- ·RAG evaluation with multilingual passages
Method
Compared against Qwen2.5-72B-Instruct and qwen-plus using identical inputs
Reviewer
CBAI Editorial Team
Plans & pricing
Self-host
Local or self-hosted use
- No public self-serve SaaS price table on the official page
Pricing may vary by region. Always verify on the vendor's website.
Feature comparison
| Feature | Qwen 1.5 | Qwen2.5-72B-Instruct | qwen-plus (hosted) |
|---|---|---|---|
| Distribution | |||
| Open weights available | |||
| Access | |||
| First-party hosted API | |||
| Modalities | |||
| Vision/audio support | |||
| Context | |||
| 32k-token window | |||
| Licensing | |||
| OSI-approved license | |||
| Ecosystem | |||
| Official quantized releases (GPTQ/AWQ/GGUF) | |||
| Runtime | |||
| Works with Ollama/llama.cpp | |||
| Support | |||
| Vendor SLA/support | |||
Is it right for you?
Good fit for
Local-first developers
Run small to medium models on a single GPU or CPU via GGUF without vendor lock-in.
Research and benchmarking
Compare sizes from 0.5B to 110B with a uniform 32k context across tasks and languages.
Fine-tuners
Train instruction or domain variants using Axolotl/LLaMA-Factory and deploy via vLLM or Ollama.
Multilingual apps
Build text-only assistants spanning 12 languages without relying on external APIs.
Less suited for
Teams needing a managed API
Alibaba retired all Qwen1.5 hosted endpoints on Aug 20, 2025; only self-hosting remains.
Vision or audio projects
Qwen 1.5 is text-only; use separate Qwen-VL or audio models instead.
Strict open-source licensing
Dense models use the custom Tongyi Qianwen license, not OSI-approved.
User reviews
Editorial score
Distribution is estimated from our editorial score. Verified user reviews coming soon.
Use cases
Typical ways teams rely on this tool — from everyday tasks to specialized workflows.
- Self-hosted chatbots
- Code generation and assistance
- Multilingual text generation
- RAG and document QA
Integrations
Reported connectors
Apps and services commonly connected out of the box or via official connectors.
- Hugging Face Transformers
- vLLM
- Ollama
Details
Category
Price
- Free to self-host / open weights
- hosted usage billed by the cloud vendor if used
Free version
Best for
- Local inference on consumer GPUs/CPUs (smaller sizes)
- Fine-tuning via Axolotl or LLaMA-Factory
- Multilingual chat and code assistants
- RAG pipelines with vLLM/Ollama
Frequently asked questions
Qwen 1.5
Developer
Ready to get started?
Visit the Qwen 1.5 website to explore plans and pricing.
Was this page helpful?
Compare Best AI may earn a commission when you click links on this page. This does not influence our editorial scores or recommendations. Advertiser disclosure