Qwen3.8 Flash

Alibaba this week unveiled Qwen3.8-Flash, an open-weight multimodal mixture-of-experts model with 125 billion parameters (6 billion active per token) offered as a prelude to Qwen4 and positioned as both a performance and value-for-money play. The Ground News aggregation of coverage from The New Stack, Intelligent Livin An Intelligent Living item in the cluster claims Qwen3.8-Flash beats DeepSeek V4 Pro on SWE-bench Pro (62.5 versus 55.4) while costing roughly a quarter of the price, and lands within striking distance of Claude Opus 5 on coding benchmarks at 33x lower input cost, framing the release as a cost-efficiency challenge to i

Alibaba's Qwen team released Qwen3.8-Flash-Next on August 26, 2026, as an open-weight multimodal mixture-of-experts model carrying 125 billion total parameters with 6 billion active per token, positioned by Alibaba as an early architectural preview of the upcoming Qwen4 family — mirroring how Qwen3-Next previewed Qwen3 On QwenCloud the model is listed as qwen3.8-flash at $0.16 per million input tokens and $0.47 per million output tokens, giving developers an aggressive price point for a long-context MoE. Alibaba's claimed benchmarks include SWE-bench Pro at 62.5 versus Claude Opus 4.6 Max at 53.4, CoWorkBench at 73.9 versus 68.2, and

AICoder reports that Alibaba Qwen open-sourced Qwen3.8-Flash, a multimodal MoE with 125B parameters plus 51B N-gram embeddings and only 6B active per token, designed as an early look at the Qwen4 architecture. The model scores 58.7 on DeepSWE 1.1, 62.5 on SWE-bench Pro, and 73.9 on CoWorkBench, beating Qwen3.7-Plus esp Weights are available on Hugging Face and ModelScope with day-0 vLLM/SGLang support, and the QwenCloud API is priced at $0.16 per million input tokens and $0.47 per million output tokens. The same launch shipped Qwen3.8-Flash-Next as an open-weights architectural preview intended to preview the Qwen4 family. Training c

Reuters reported that Alibaba's Qwen released Qwen3.8-Flash, a new multimodal model delivering stronger coding and office-task performance at substantially reduced training costs. Qwen3.8-Flash supports a default context window of 262,144 tokens, expandable to 1 million tokens, enabling it to handle large files, length For API access, Qwen said it would charge 1 yuan ($0.1488) per million input tokens and 3 yuan per million output tokens, aligning with the Model Studio/Qwen Cloud pricing announced alongside the launch. Alibaba additionally released open-source weights for Qwen3.8-Flash-Next, described by Qwen as a prototype architect

Alibaba's Qwen team released Qwen3.8-Flash on August 26, 2026, positioning it as a multimodal model with stronger coding and office-task performance while cutting training costs to roughly one-ninth of those required for the previous Qwen3.7-Plus model, according to a Qwen statement carried by Reuters. The launch sits Qwen3.8-Flash supports a default 262,144-token context window that can be extended to 1 million tokens for handling large files, lengthy conversations, and extensive research material, Reuters reported. Pricing via Alibaba's API is set at 1 yuan (about US$0.1488) per million input tokens and 3 yuan per million output t

Alibaba officially unveiled Qwen3.8-Flash as an open-weight, multimodal Mixture-of-Experts (MoE) model designed for an optimal balance of capability, latency, and cost. The architecture pairs a 125B-parameter main model with an additional 51B N-gram embeddings component, activating only 6B parameters per token. Qwen3.8 Qwen3.8-Flash demonstrates competitive agentic coding, long-horizon agent, and multimodal performance against leading models including DeepSeek-V4-Flash and Claude-Opus-4.6 across benchmarks such as SWE-bench Pro, CoWorkBench, Toolathlon Verified, MathVision, AndroidWorld, and ERQA. API access is priced at 1 RMB (~$0.1