Qwen: Qwen3 8B

Qwen3-8B is a dense 8.2B parameter causal language model from the Qwen3 series, engineered for both rigorous reasoning tasks and efficient dialogue. It features a unique ability to seamlessly switch between a "thinking" mode for math, coding, and logical inference, and a "non-thinking" mode for general conversation, making it highly adaptable. This model is fine-tuned for superior instruction-following, agent integration, and creative writing. It offers robust multilingual support across 100+ languages and dialects, natively handles a 32K token context window, and can extend up to 131K tokens with YaRN scaling. Capabilities include functions, code generation, and streaming. It's best for chat, code, and math tasks. Limitations include no image generation and no internet access. Pricing is $0.05/0.25 per 1M tokens (input/output), available for FREE access.