References 685% CONFIDENCEOverall confidence: 85%How well the pin's source and references back up its dates.Weighted average of how firmly 6 references, the source included, support the pin's start and end times; a reference counts half as much for every 180 days older than the newestShow all pins at 75% confidence or better

The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.

• [1]90%
  huggingface.co/Qwen/Qwen3-235B-A22B-Thinking-2507

  huggingface.co· Posted Oct 12, 2026· Starts Jul 25, 2025· 28% of score

  VentureBeat's^[4] 25 July 2025 article reports the release that day; the Qwen3 repository's news log reads "2025.07.25: We released the updated version of Qwen3-235B-A22B thinking mode, named Qwen3-235B-A22B-Thinking-2507", and OpenRouter^[6] lists it as released Jul 25, 2025.

• [2]75%
  Qwen3 235B A22B 2507 (Reasoning) - Artificial Analysis

  artificialanalysis.ai· Added Oct 12, 2026· 28% of score

  Artificial Analysis: Qwen3 235B A22B 2507 (Reasoning), open weights, released July 2025, 256k context, scores 13 on its current Intelligence Index (median for comparable models 18), ranked 83 of 117, and is flagged deprecated in favour of Qwen3.5 397B A17B.

• [3]90%
  QwenLM/Qwen3 - GitHub

  github.com· Added Oct 12, 2026· 28% of score

  Qwen's repository news log dates the release 2025.07.25 and records the 8 August 2025 update that lets the 2507 models handle 1 million tokens; it also lists Qwen3-235B-A22B-Instruct-2507 (21 July) and the 30B-A3B 2507 models (30 and 31 July).^[1]

• [4]92%
  It's Qwen's summer: new open source Qwen3-235B-A22B-Thinking-2507 tops OpenAI, Gemini reasoning models on key benchmarks - VentureBeat

  venturebeat.com· Published Jul 25, 2025· Starts Jul 25, 2025 ✓· 5% of score

  VentureBeat, 25 July 2025: Alibaba's Qwen Team released Qwen3-235B-A22B-Thinking-2507 under Apache 2.0; AIME25 92.3, LiveCodeBench v6 74.1, Arena-Hard v2 79.7; Qwen moving to separate reasoning and instruct models instead of a hybrid mode; Alibaba API $0.70 input and $8.40 output per million tokens.

• [5]85%
  Qwen3-235B-A22B-Thinking-2507 - Simon Willison, 25 July 2025

  simonwillison.net· Published Jul 25, 2025· 5% of score

  Simon Willison: the third Qwen release of the week after Qwen3-235B-A22B-Instruct-2507 (Monday 21st) and Qwen3-Coder-480B-A35B-Instruct (Tuesday 22nd); an FP8 variant too; context raised to 262,144 tokens from April's 32,768 native; on OpenRouter^[6] the same day, where it thought for 166 seconds on a pelican-on-a-bicycle SVG.