Alibaba debuts Qwen3.5, a 397B-parameter open-weight multimodal AI model that it says is 60% cheaper to use and 8x better at large workloads than Qwen3
ReutersEduardo Baptista
Context & Ripple Effects
Alibaba had already established Qwen as an open-weight reasoning-model family with its Qwen3 hybrid reasoning models. Qwen3.5 extends that line into a much larger multimodal release while making an explicit cost-and-large-workload performance claim.
The significance is less the parameter count alone than the combination of open weights, multimodal capability, and Alibaba’s stated operating-efficiency gains. It gives model adopters another option to evaluate on cost per useful workload, not just benchmark quality.
First-order effects
Developers and enterprises able to deploy open-weight models can test Qwen3.5 for multimodal workloads, with Alibaba claiming 60% lower use costs and substantially stronger performance on large workloads versus Qwen3.
Alibaba strengthens the Qwen portfolio’s positioning by pairing a 397B-parameter model with an efficiency claim, rather than treating larger scale as the sole product advance.
Second-order effects
Other open-weight model suppliers face more pressure to show both deployment economics and multimodal capability, not merely release larger models.
Buyers evaluating proprietary and open-weight systems gain leverage in procurement: a credible lower-cost alternative can shift evaluations toward workload-specific price-performance comparisons.
Third-order effects
If releases like Qwen3.5 deliver on their claims in real deployments, competition may increasingly center on inference efficiency and integration support, compressing the value of raw model scale alone.
Open-weight multimodal models could broaden the surrounding market for hosting, optimization, and enterprise deployment services, while making model selection a more active buyer decision.
The trend: AI model competition is shifting toward open-weight, multimodal systems differentiated by cost per useful workload as much as by frontier scale.
A clarification of Qwen3.5 Plus and 397B: 1. for opensource, we follow the tradition to make parameters apparent so we use the name with the number of total parameters and active params. 2. Qwen3-Plus is a hosted API version of 397B. As the model natively supports 256K tokens, …
🚀 Qwen3.5-397B-A17B is here: The first open-weight model in the Qwen3.5 series. 🖼️Native multimodal. Trained for real-world agents. ✨Powered by hybrid linear attention + sparse MoE and large-scale RL environment scaling. ⚡8.6x-19.0x decoding throughput vs Qwen3-Max 🌍201 langu…
Let's do the KV cache math for Qwen3.5: - KV heads: 2 - Head dimension: 256 - gated attention layers: 15 - bytes per element (BF16): 2 2 x 256 x 15 x 2 = 15 360 This is the same for K and V. So, we multiply by 2: 30 720 bytes Roughly 31 kb per token of context. Meaning at max
Qwen3.5 runs quite well in mlx-lm. Awesome that we have a frontier-level hybrid model. The context gets longer but the inference speed and memory use barely change. Here's the Q4 generating a space invaders game on an M3 Ultra. Generated 4,120 tokens at 37.6 tok/s. [video]
🎊 Kudos to the teams at @Alibaba_Qwen on the launch of Qwen 3.5 with Qwen3.5-397B-A17B. 🙌 Developers can start building today for free: https://build.nvidia.com/... Or download and customize it with NVIDIA NeMo: https://github.com/... [video]
The new @Alibaba_Qwen Qwen3.5-397B-A17B is live on OpenRouter now! This multimodal model uses a hybrid architecture combining linear attention with sparse MoE for higher inference efficiency. Available as both the open weights version and Qwen3.5 Plus with extended 1M context.
Qwen 3.5 has been released. 397B parameters including 17B active ones. This is the flagship open sourced version. Looks to be reasonably better than Qwen3 235B. I've found larger Qwen models to not be all that great. Given the recent Minimax & GLM releases, Qwen has to also have …
🎉 Meet Qwen3.5-397B-A17B from @Alibaba_Qwen, 397B total params (17B active), built for real-world multimodal intelligence — day-0 support is now live in SGLang! 👁️ Unified vision-language foundation (early fusion): stronger reasoning, coding & agents ⚡ Gated DeltaNet + sparse [im…
The new chonky Qwen 3.5 looks pretty solid, beating their own Qwen3-Max model everywhere and is much better at vision benchmarks than Qwen3-235B-A22B-VL Now what I sadly haven't seen is anything on reasoning efficiency. [image]
You can now run Qwen3.5 locally! 💜 Qwen3.5-397B-A17B is an open MoE vision reasoning LLM for agentic coding & chat. It performs on par with Gemini 3 Pro, Claude Opus 4.5 & GPT-5.2. Run 4-bit on 256GB Mac / RAM. Guide: https://unsloth.ai/... GGUF: https://huggingface.co/... [image…
Qwen3.5 is Live! Today we openweight the first model, Qwen3-397B-A17B, which is a native multimodal model supporting both thinking and non-thinking modes. We have strengthened its coding and agentic capabilities to foster productivity for developers and enterprises. Hope you
So speaking of benchmarks, what can be said of the new open Qwen? First, it completely destroys Qwen3-VL-235B ofc, but more surprisingly it outscores Qwen3-Max-thinking. All the while it's the same model as “Plus”. Plus just has 1M context and some more bells and whistles. [image…
BREAKING Qwen3.5-397B-A17B weights have been uploaded to Hugging Face > 397B parameters in total > 17B active parameters per token > 256K context window > expendable to 1M token more soon once i am done running my evals [image]