# Qwen3.8-Flash-Next Qwen3.8-Flash-Next is an AI model released by Qwen on Aug 26 2026. It has 125B parameters, a 262k context window and open weight. At release it scored 48.1% on NL2Repo-Bench, 55.7% on JobBench and 73.9% on CoWorkBench. ## Facts | Field | Value | | --- | --- | | Model | Qwen3.8-Flash-Next | | Developer | Qwen | | Release date | Wednesday, Aug 26 2026 | | Licensing | Open Weight | | Parameters | 125B | | Context window | 262k | ## Benchmark scores published at release | Benchmark | Score | Source | What it measures | | --- | --- | --- | --- | | SWE-Bench Pro | 62.5% | Lab | Can the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. | | SWE-Bench Multilingual | 81% | Lab | Like SWE-Bench, but the coding problems span many programming languages, not just one. Tests how broadly the AI can code. Higher is better. | | DeepSWE 1.1 | 58.7% | Lab | Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better. | | NL2Repo-Bench | 48.1% | Lab | Tests whether the AI can turn a natural-language requirement into working code across an entire repository, not just produce a single function or patch. Higher is better. | | LiveCodeBench | 91.9% | Lab | Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | | JobBench | 55.7% | Lab | Tests the AI on professional workplace tasks that require using real work tools — the kind of multi-step jobs an office worker handles. Higher is better. | | CoWorkBench | 73.9% | Lab | Tests long-running office tasks across fields including computer science, finance, law, medicine, and other productivity work. Higher is better. | | Toolathlon-Verified | 73.5% | Lab | Tests how well the AI uses everyday personal tools and apps to get things done — a human-checked version of Toolathlon. Higher is better. | | Humanity's Last Exam (no tools) | 35.9% | Lab | Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. | | GPQA Diamond | 91.7% | Lab | Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | | IFBench | 81.3% | Lab | Tests whether the AI can follow detailed instructions and satisfy multiple constraints at once. Higher is better. | | OSWorld 2.0 | 19.4% | Lab | Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Version 2.0 is a harder, refreshed task set. Higher is better. | | Agent's Last Exam (pass@1) | 24.3% | Lab | A hard set of desktop and operating-system tasks an AI agent has to finish by looking at the screen and working the machine itself. The score is the share it passes outright — partial credit does not count. Higher is better. | | Agent's Last Exam (score) | 51.2% | Lab | The graded score on the same desktop and operating-system tasks in Agent's Last Exam, giving partial credit for progress beyond the strict pass-or-fail result. Higher is better. | | CharXiv Reasoning | 84.6% | Lab | Can the AI read and reason about complex charts and figures, not just text? Higher is better. | | LVBench | 76.6% | Lab | Can the AI follow a very long video — up to an hour — and answer questions that need details from far apart in it? Higher is better. | ## About Qwen3.8-Flash-Next Qwen3.8-Flash-Next, released August 26, 2026, was Alibaba's public preview of the architecture it intended to build Qwen4 on, put out as downloadable weights under a custom qwen-community licence. It was a natively multimodal mixture-of-experts model with 125 billion parameters and only 6 billion active per token, carrying a further 51 billion parameters of N-gram embeddings — a way of adding capacity that costs less computation than adding experts and is easier to offload from accelerator memory. Its 262K-token native context window extended to roughly 1M through YaRN. The rest of the design was equally experimental. Gated DeltaNet was paired with Qwen Sparse Attention, which selects context at the level of micro-blocks rather than individual tokens to cut latency on the long agentic prompts Alibaba said were coming to dominate real usage; a gated residual stream added separate read and write gates; and the training recipe split the Muon and AdamW optimisers across different weight categories and started at the target batch size instead of warming up to it. Alibaba put the training cost at roughly one-ninth that of Qwen3.7-Plus, the 397B proprietary model it was benchmarked against. The launch table was built to show a 6B-active model beating much larger ones. It scored 62.5% on SWE-bench Pro and 81.0% on SWE-bench Multilingual, 58.7% on DeepSWE 1.1, 73.9% on CoWorkBench and 55.7% on JobBench — the last of those 19 points above the 36.6% Alibaba reported for Claude Opus 4.6 Max and 28 above Qwen3.7-Plus. It posted 91.7% on GPQA Diamond and 91.9% on LiveCodeBench v6, and 35.9% on HLE, the one language row where Opus 4.6 Max led at 40.0%. Computer use was the visible gap: 19.4% on the binary scoring of OSWorld 2.0, level with the smaller Qwen3.8-27B. Alongside the weights Alibaba announced a hosted production model, Qwen3.8-Flash, at $0.16 per million input tokens and $0.47 per million output on its QwenCloud API. ## Questions and answers ### When was Qwen3.8-Flash-Next released? Qwen3.8-Flash-Next was released by Qwen on Wednesday, Aug 26 2026. ### Who made Qwen3.8-Flash-Next? Qwen3.8-Flash-Next was built by Qwen. Alibaba's AI lab, building the Qwen family. The most prolific publisher of open-weight models of any major lab, alongside a proprietary Max and Plus tier sold through Alibaba Cloud. ### What benchmark scores did Qwen3.8-Flash-Next get? Qwen3.8-Flash-Next reports 16 tracked benchmark scores — SWE-Bench Pro: 62.5%; SWE-Bench Multilingual: 81%; DeepSWE 1.1: 58.7%; NL2Repo-Bench: 48.1%; LiveCodeBench: 91.9%; JobBench: 55.7%; CoWorkBench: 73.9%; Toolathlon-Verified: 73.5%; Humanity's Last Exam (no tools): 35.9%; GPQA Diamond: 91.7%; IFBench: 81.3%; OSWorld 2.0: 19.4%; Agent's Last Exam (pass@1): 24.3%; Agent's Last Exam (score): 51.2%; CharXiv Reasoning: 84.6%; LVBench: 76.6%. Scores are the figures published at release by Qwen. It holds the best score among all models tracked here on NL2Repo-Bench, JobBench, CoWorkBench, IFBench and Agent's Last Exam (score). ### What is the context window of Qwen3.8-Flash-Next? Qwen3.8-Flash-Next has a context window of 262k. That is the maximum amount of input plus output the model can hold in a single request. ### How many parameters does Qwen3.8-Flash-Next have? Qwen3.8-Flash-Next is reported at 125B parameters. ### Is Qwen3.8-Flash-Next open source? Partly. Qwen3.8-Flash-Next is an open-weight model: the trained weights are free to download and run locally or on your own infrastructure, but the training data and code are not fully released and the license may restrict some commercial uses. It is not open source in the strict sense. ### What came before and after Qwen3.8-Flash-Next? Qwen's previous tracked release was Qwen3.8-27B on Aug 14 2026, 12 days earlier. It is the most recent Qwen model tracked on AI Release Tracker. --- Canonical page: https://aireleasetracker.com/model/qwen/qwen3.8-flash-next Full dataset: https://aireleasetracker.com/llms-full.txt · JSON: https://aireleasetracker.com/models.json Source: AI Release Tracker (https://aireleasetracker.com). Benchmark scores are the figures published by the releasing lab at launch.