Alibaba

Models & products

Models

Qwen

Trends

• Chinese labs take the lead in open-weight models
• Open weights consolidate around a flash tier and an Nvidia-owned hub
• Voice AI splits into commodity APIs and reasoning-grade agents
• Agents improve themselves through skills, harnesses, and memory

Items

1. Oct 2

   Tsinghua and Alibaba's PoS belief states

   An inference-time framework that builds and continually validates explicit belief states combining current-world estimates with unresolved task requirements, detecting "Belief Trapping" and recovering from it to top four long-horizon agent benchmarks across three LLM backbones.

2. Sep 18

   Alibaba's Qwen3.8-Omni-Flash

   Alibaba's first omni-modal model built for agentic use takes text, image, audio, and video with a 1M-token context, scores more than 25% above Qwen3.5-Omni-Plus across 29 benchmarks, and is API-only at $0.15/$0.47 per 1M tokens.

3. Sep 4

   Qwen's Terminal-Universe rebuilds agent workspaces from trajectories

   Instead of building agent-training environments from scratch, the framework replays tool-execution histories to reconstruct 37.3k executable workspaces, lifting Qwen3.5-27B by 11.9 points on Terminal-Bench 2.1 and 13.8 on multi-round EvoCode-Bench v2 after fine-tuning.

4. Aug 26

   Alibaba's Qwen3.8-Flash-Next

   An explicit Qwen4-architecture preview pairing 125B parameters with a 51B N-gram "engram" lookup table at only 6B active per token, released with open weights on Hugging Face and ModelScope and priced roughly a third below DeepSeek-V4-Flash.

5. Aug 15

   Alibaba's Qwen3.8-27B ties GPT-5.6 Luna

   The dense, natively multimodal 27B model (Apache 2.0, 262K context) that completes the Qwen3.8 open-weight rollout scored 52 on the Artificial Analysis Intelligence Index in independent testing, tying hosted GPT-5.6 Luna from a roughly 17GB file on one consumer GPU.

6. Aug 13

   Alibaba releases Qwen3.8-2.4T-A95B open weights

   Alibaba shipped the weights it promised the week before, its first Max-class open release: a 2.4T-parameter (95B active) MoE with a 256K native context extensible to about 1M, on Hugging Face and ModelScope alongside an FP8 build.

7. Aug 8

   Alibaba's DreamX-Phi 1.0 manipulation world model

   An action-conditioned video world model that injects per-arm SE(3) transforms through geometric attention, plus a depth branch and object masks, to predict faithful manipulation rollouts, taking first place on Track 1 of the WorldArena 2.0 Challenge.

8. Aug 3

   Alibaba's Qwen3.8-Max reaches general availability

   The 2.4T-parameter (95B active) multimodal MoE with a 1M-token context went GA via API at $2/$6 per 1M tokens, and Alibaba committed to open weights for it and a smaller Qwen3.8-27B, its first open-weight Max-class model.

9. Jun 23

   Alibaba's Qwen-AgentWorld world models

   The first open-weight language world models (35B-A3B and 397B-A17B) simulate agentic environments across seven domains via long chain-of-thought, serving as decoupled simulators that improve agentic RL beyond real-environment training, released with the AgentWorldBench evaluation.

10. May 20

    Alibaba's Qwen3.7-Max

    An API-only model built for long-horizon autonomy ran 35 hours of unsupervised kernel optimization (1,158 tool calls) for a 10x speedup versus DeepSeek V4 Pro's 3.3x, and scores 80.4% on SWE-Verified and 92.4% on GPQA Diamond.