Qwen3-8B
Efficient dense 8B with Qwen3's hybrid thinking, sized for single-GPU use
About
Qwen3-8B brings the Qwen3 family's hybrid thinking/non-thinking design down to 8 billion dense parameters, released April 29, 2025 under Apache 2.0. It keeps the 128K context window and the multilingual coverage of its larger siblings while fitting comfortably on a single consumer GPU once quantized. The team reports it punching above its weight on reasoning and multilingual tasks for the size class. It found its niche in local chatbots, RAG pipelines, and agent prototypes where 32B is too heavy and 4B too thin. Like the rest of Qwen3 it is widely mirrored across Hugging Face, ModelScope, and Kaggle with broad inference-stack support.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: strong performance for its size class on reasoning and multilingual benchmarks (team-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Released 2025-04-29 alongside Qwen3 family, per Alibaba; Apache 2.0 confirmed; Dense 8B, hybrid thinking modes, 128K context (team specs); Single-consumer-GPU fit when quantized; broad inference support.
More in the qwen3 family
- Qwen3-32B-AWQ
- Qwen3-Next-80B-A3B
- Qwen3-235B-A22B
- Qwen3-235B-A22B-2507
- Qwen3-4B-Instruct-2507
- Qwen3.5
- Qwen3-235B-A22B-Instruct-2507
- Qwen3 235B
- Qwen3-235B-A22B-Thinking-2507
- Qwen3-Omni-Flash
- Qwen3.6-27B
- Qwen3-32B
- Qwen3-Max
- Qwen3.8-Flash-Next
- Qwen3-Coder-Plus
- Qwen3
- Qwen3-0.6B
- Qwen3-4B
- Qwen3-Coder-Next
- Qwen3-4B-2507
- Qwen3-14B
- Qwen3-VL-30B-A3B
- Qwen3-1.7B
- Qwen3-Omni-30B-A3B