{"slug": "alibaba-launches-qwen-3-8-with-2-4-trillion-parameters-claims-near-frontier", "title": "Alibaba Launches Qwen 3.8 With 2.4 Trillion Parameters, Claims Near-Frontier Performance", "summary": "Alibaba's Qwen team on July 19 unveiled Qwen 3.8, a 2.4 trillion-parameter multimodal AI model that the company says is 'second only to Fable 5' — a reference to Anthropic's top-tier Claude model — though Alibaba has not published independent benchmarks to support the claim. The model, built on a sparse Mixture-of-Experts architecture, is available in preview through Alibaba's Token Plan, Qoder, and QoderWork at 10% of standard pricing, with an open-weight release promised 'soon.'", "body_md": "# Alibaba Launches Qwen 3.8 With 2.4 Trillion Parameters, Claims Near-Frontier Performance\n\n- Qwen 3.8 contains 2.4 trillion parameters, making it the second-largest publicly known model behind Moonshot AI's Kimi K3 at 2.8 trillion\n[[1]](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/) - Alibaba claims the model is 'second only to Fable 5' but has not published independent benchmarks to verify that ranking\n[[2]](https://officechai.com/ai/alibaba-qwen-3-8/) - The model is available in preview through Alibaba's Token Plan, Qoder, and QoderWork at 10% of standard pricing\n[[1]](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/) - An open-weight release is promised 'soon,' which would break Alibaba's pattern of keeping Max-tier models closed-source\n[[3]](https://www.buildfastwithai.com/blogs/qwen3-8-preview-2-4t-params-open-weights-release) - Alibaba (BABA) shares closed at $114.97 on July 17, giving the company a market cap of roughly $262 billion\n[[4]](https://stockanalysis.com/stocks/baba/)\n\nAlibaba's Qwen team on July 19 unveiled Qwen 3.8, a 2.4 trillion-parameter multimodal AI model that the company says rivals the world's leading frontier systems. The team described the model as \"second only to Fable 5\" — a reference to Anthropic's top-tier Claude model — though Alibaba has not yet published independent benchmarks to support the claim [1].\n\nQwen 3.8 is built on a sparse Mixture-of-Experts architecture and represents the team's first multimodal model exceeding one trillion parameters. It can process text, images, videos, and documents. A preview is available now through Alibaba's Token Plan subscription service, its Qoder coding platform, and QoderWork productivity suite at 10% of standard pricing [[1]](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/) [2].\n\nThe launch escalates an already fierce parameter race among Chinese AI labs. Moonshot AI — in which Alibaba holds a 36% stake — released Kimi K3 at 2.8 trillion parameters earlier this year, briefly holding the title of the largest open-source model. Qwen 3.8 slots in as the second-largest publicly disclosed model [3].\n\n## What's in the Model\n\nQwen 3.8 succeeds Qwen 3.7-Max and represents a significant jump in scale. For reference, Qwen 3.7-Max scored 92.4% on GPQA Diamond, 80.4% on SWE-bench Verified, and 69.7% on Terminal-Bench 2.0 [3]. Alibaba has not disclosed how many of Qwen 3.8's 2.4 trillion parameters are active at inference time — a critical detail for sparse MoE models, where only a fraction of parameters are engaged per query.\n\nThe Qwen team said the model should outperform its predecessor in coding, full-stack development, data analysis, and office workflows [1]. Developer Shuai Bai described it as 'the team's first multimodal model with more than 1 trillion parameters' capable of handling multiple content types\n\n.\n\n[[1]](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/)## Pricing and Access\n\nThe preview is bundled into Alibaba's Token Plan subscription tiers. The Lite plan runs $6 for 2,500 credits per week, while the Pro tier costs $68 for 40,000 weekly credits and supports six to eight concurrent agents [2]. Both tiers offer compatibility with OpenAI and Anthropic API protocols, allowing developers to integrate Qwen 3.8 without restructuring existing workflows.\n\nAlibaba has not disclosed per-token pricing for standalone API access. For comparison, the predecessor Qwen 3.7-Max was priced at $1.25 per million input tokens and $3.75 per million output tokens [3].\n\n## The Open-Weight Question\n\nAlibaba promised an open-weight release 'soon,' but has not committed to a date [1]. If the company follows through, Qwen 3.8 would be by far the largest open-weight model ever released — surpassing even Kimi K3's open-source offering at 2.8 trillion parameters in strategic significance, given Alibaba's broader developer ecosystem.\n\nHowever, Alibaba's Max-tier models have historically remained closed-source, making the open-weight pledge a departure from established practice [3]. The distinction matters: open-weight availability would allow developers to self-host, fine-tune, and modify the model without relying on Alibaba's cloud infrastructure.\n\n## Competitive Landscape\n\nThe launch arrives amid an intense period for Chinese AI. Moonshot AI's Kimi K3, released earlier this year at 2.8 trillion parameters, shook global technology stocks and challenged assumptions about the gap between Chinese and Western frontier models. Zhipu AI's GLM 5.2 added further competitive pressure [2].\n\nAlibaba's broader AI strategy extends beyond Qwen itself. The company recently secured a role as Apple Intelligence's technology partner in China after Beijing approved the service [5]. That deal gives Alibaba distribution reach into Apple's massive Chinese user base — a commercial advantage that complements the Qwen model lineup.\n\nWithout published benchmarks, Alibaba's claim of near-frontier performance remains unverified. The AI community will be watching for third-party evaluations and the promised open-weight release to assess whether Qwen 3.8 truly narrows the gap with Anthropic's Claude Fable 5 and OpenAI's latest models.\n\n## Market Context\n\nAlibaba shares closed at $114.97 on July 17, down 2.14% on the session, giving the company a market capitalization of approximately $262 billion [4]. The stock has been volatile in recent months, posting a 12% rally over two days earlier in July on a combination of a DOJ settlement and cloud-related catalysts\n\n.\n\n[[6]](https://www.tradingkey.com/analysis/stocks/us-stocks/262022410-alibaba-baba-stock-forecast-july-2026-doj-settlement-cloud-h200-tradingkey)The Qwen 3.8 announcement landed after U.S. market hours on a Saturday, meaning any investor reaction will materialize when trading resumes Monday.\n\n## Companies mentioned\n\n## Further sources\n\n[[1] The Decoder — Alibaba's Qwen takes on Kimi K3 with open-weight Qwen 3.8, says m… ↗](https://the-decoder.com/alibabas-qwen-takes-on-kimi-k3-with-open-weight-qwen-3-8-says-model-is-second-only-to-fable-5/)\n\n[[2] OfficeChai — Alibaba Announces 2.4 Trillion-Parameter Open-Weight Qwen 3.8, Say… ↗](https://officechai.com/ai/alibaba-qwen-3-8/)\n\n[[3] BuildFastWithAI — Qwen3.8 Preview: 2.4T Params, Open Weights, Release ↗](https://www.buildfastwithai.com/blogs/qwen3-8-preview-2-4t-params-open-weights-release)\n\n[[4] Stock Analysis — Alibaba Group Holding (BABA) Stock Price & Overview ↗](https://stockanalysis.com/stocks/baba/)\n\n[[5] Yahoo Finance — Alibaba's Qwen Unveils Preview of Flagship AI Model ↗](https://finance.yahoo.com/technology/ai/articles/alibaba-qwen-unveils-preview-flagship-110209258.html)\n\n[[6] TradingKey — Alibaba (BABA) Stock Forecast July 2026 ↗](https://www.tradingkey.com/analysis/stocks/us-stocks/262022410-alibaba-baba-stock-forecast-july-2026-doj-settlement-cloud-h200-tradingkey)\n\nThe stories that matter, in one email. Free — unsubscribe anytime.", "url": "https://wpnews.pro/news/alibaba-launches-qwen-3-8-with-2-4-trillion-parameters-claims-near-frontier", "canonical_source": "https://mlq.ai/news/alibaba-launches-qwen-38-with-24-trillion-parameters-claims-near-frontier-performance/", "published_at": "2026-07-19 18:05:59+00:00", "updated_at": "2026-07-20 12:58:25.433999+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "generative-ai", "ai-products", "ai-startups"], "entities": ["Alibaba", "Qwen 3.8", "Moonshot AI", "Kimi K3", "Anthropic", "Claude", "Qoder", "Token Plan"], "alternates": {"html": "https://wpnews.pro/news/alibaba-launches-qwen-3-8-with-2-4-trillion-parameters-claims-near-frontier", "markdown": "https://wpnews.pro/news/alibaba-launches-qwen-3-8-with-2-4-trillion-parameters-claims-near-frontier.md", "text": "https://wpnews.pro/news/alibaba-launches-qwen-3-8-with-2-4-trillion-parameters-claims-near-frontier.txt", "jsonld": "https://wpnews.pro/news/alibaba-launches-qwen-3-8-with-2-4-trillion-parameters-claims-near-frontier.jsonld"}}