Qwen

Alibaba open-weights Qwen3.8-Flash-Next, a 125B MoE with 6B active parameters and 51B N-gram embedding parameters. It uses Gated DeltaNet plus sparse attention, previews the Qwen4 architecture, and supports 262K native context extensible to 1M. API pricing is $0.16/$0.47 per million tokens.

Alibaba released Qwen3.8-27B as open weights with day-0 SGLang support: 206 tok/s decode on a single RTX 5090 using NVFP4 with DSpark, and 38 tok/s on a DGX Spark — a small dense model aimed squarely at running on hardware people already own.

Alibaba refreshed its Qwen3.8-Max snapshot: 2.4T parameters, a 1M token context window, and extra post training for coding and long horizon workflows. Live on QwenCloud at $2 per million input tokens and $6 output, with explicit cache hits at $0.17 and implicit at $0.25.

Qwen3.8-Max, better and cheaper. Try it out. 👀 https://t.co/zHXKtgZRci

Qwen announced free access to Qwen3.8-Max, Qwen3.8-Flash and the Wan3.0 video model through GMI Cloud for a limited week, with rate limits raised across all three models.

Qwen3.8-27B is now packaged for LM Studio and needs roughly 17GB to run locally, putting a fresh open-weights release on consumer hardware without an API bill. Practitioners in the replies report multi-hundred-token/s prefill and long-context runs on Apple silicon.

Quantization and attention settings for serving Qwen 3.8 27B, published as recipes.

Alibaba's Qwen3.8-Flash is now available as a model option in OpenCode Go: 125B total parameters with 6B active, a 1M-token context window, and multimodal input. Sparse activation at that size is the pitch — long-context agent runs at small-model inference cost.

Arena's Image-to-WebDev leaderboard puts Qwen3.8-27B first among open models and seventh overall at 1574 points, priced at $0.40/$3 per million tokens — screenshot-to-frontend work from a 27B model, with another Qwen release teased the same day.

Qwen3.8-27B lands on Ollama, pitched as delivering consistently across whatever harness you run it in.

Qwen's team benchmarks 18 frontier models running simulated online stores for 365 days against 576 suppliers, 152 of them fraudulent. Deterministic negotiation makes results reproducible. No model leads all seven axes, and most fail to learn cheaper buying over the year.

Alibaba's Qwen3.8-Flash is now served through OpenRouter — a multimodal reasoning model positioned for coding assistants, agentic workflows, codebase and document analysis, desktop interaction, charts, and long-video understanding.

Qwen's new Flash tier is priced for volume: $0.15 per million input tokens, $0.47 output, and $0.016 on cache hits, with 262K native context extensible to 1M. A cheap default for long-context agent loops where cache reads dominate the bill.

Qwen-Image-2.1, Alibaba's unified image generation and editing checkpoint, ships with day-zero OpenVINO optimization from Intel, letting developers run it efficiently on Intel CPUs and NPUs right away.

Qwen3.8-27B is now served on Cerebras hardware for fast inference, with Artificial Analysis scoring it near GPT-5.6, DeepSeek V4 Pro, and Claude Sonnet 4.6, though early testers report it underperforms on coding tasks.

Alibaba Qwen's QwenGyre framework tackles online RL for hours-long agent runs. An elastic scheduler moves GPUs between rollout and training without killing live executions, and a trajectory processor turns compaction, sub-agent and retry branches into deduplicated prefix-shared trees.

Qwen released Audio 3.1: upgraded ASR, TTS and full-duplex Realtime models, plus ASR-Next and TTS-Next. Pricing is cut by up to 95%. ASR and Realtime APIs are live on Qwen Cloud, and the other APIs are coming soon. TTS-Next and ASR-Next claims are not yet independently documented.

Alibaba's Qwen team ships an omni-modal model pairing native audio-video understanding with agentic tool use, cutting video input costs ~89% versus the prior Omni model and adding a 1M-token context with agentic video search.

Alibaba's Qwen Intelligence ships planner, mobile-use and creative agents. The planner paper reports 77.05% on MobilePA-Bench and the UI agent 82.1% on MobileWorld, all self-reported. Alibaba also opens MobileWorld and MobilePA-Bench; the planner repo is a report site, not weights or code.

Alibaba's Qwen team released Qwen-Image-2.1-Turbo, an 8-step accelerated 7B checkpoint for text-to-image and editing. It loads via a Diffusers pipeline with a bundled sampling schedule and uses CFG=1 with prefix KV caching. Pro and Turbo hosted APIs launched alongside. License not verified.

Qwen-Image-2.1 ranked #1 among open-source models in both the Image Edit and Text-to-Image Arenas, scoring 1367 points and landing just 3 points behind GPT-Image-1.5-high-fidelity overall.

Qwen's newest simultaneous interpretation model uses an Interleave architecture to cut average lag from 2.8s to 2.3s across 60 languages, adding real-time speaker diarization with voice cloning and synced bilingual captions.

NVIDIA NeMo AutoModel adds day-0 supervised fine-tuning for Qwen3.8-Flash-Next, a hybrid-attention MoE with an Engram n-gram memory table. The doc covers FSDP2, expert and context parallelism, numerical validation against SGLang, and limits such as frozen QSA indexers.