Jin10 data July 22 news, Tongyi Qianwen: Today, we have updated the flagship Qwen3 model and launched the updated version of Qwen3-235B-A22B-FP8 in non-thinking mode, named Qwen3-235B-A22B-Instruct-2507-FP8. The new Qwen3 model shows significant improvements in general capabilities, including instruction following, logical reasoning, text comprehension, mathematics, science, programming, and tool usage. It has performed exceptionally well in various evaluations such as GQPA (knowledge), AIME25 (mathematics), LiveCodeBench (programming), Arena-Hard (human preference alignment), and BFCL (Agent capability), surpassing top open-source models like Kimi-K2 and DeepSeek-V3, as well as leading closed-source models like Claude-Opus4-Non-thinking.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
Tongyi Qianwen: Update flagship version Qwen3 model, universal capabilities significantly enhanced.
Jin10 data July 22 news, Tongyi Qianwen: Today, we have updated the flagship Qwen3 model and launched the updated version of Qwen3-235B-A22B-FP8 in non-thinking mode, named Qwen3-235B-A22B-Instruct-2507-FP8. The new Qwen3 model shows significant improvements in general capabilities, including instruction following, logical reasoning, text comprehension, mathematics, science, programming, and tool usage. It has performed exceptionally well in various evaluations such as GQPA (knowledge), AIME25 (mathematics), LiveCodeBench (programming), Arena-Hard (human preference alignment), and BFCL (Agent capability), surpassing top open-source models like Kimi-K2 and DeepSeek-V3, as well as leading closed-source models like Claude-Opus4-Non-thinking.