Qwen3.5-4B
Language Model by Alibaba - Research mention via arxiv
This page stays evidence-first until LM Market Cap can verify real availability, pricing, and benchmark coverage.
Source Evidence
The method, which we call S0 tuning, optimizes one state matrix per recurrent layer while freezing all model weights. On Qwen3.5-4B (GatedDeltaNet hybrid), S0 tuning improves greedy pass@1 by +23.6 +/- 1.7 pp (10 seeds).
Detected Signals
Resources
Verification Status
This page tracks a pre-release signal for Qwen3.5-4B. Once availability is confirmed, it will appear on our LLM Leaderboard with full scoring, benchmarks, and pricing. Until then, release date, pricing, and benchmark claims remain unconfirmed unless they are directly cited above. Check our New AI Models page for the latest releases and the source links above for the current evidence.
More from Alibaba
Qwen3.5-4B was first detected on April 2, 2026 via arxiv. That is a detection signal, not a confirmed launch date, so we keep tracking the source evidence until official availability is verified.
Not yet. The current evidence comes from repository, research, or other detection signals rather than a confirmed provider launch post. Treat this page as a watch page until an official announcement appears.
You can verify the source links, the detection dates, and the evidence type shown on this page. We do not treat release date, pricing, or benchmark availability as confirmed unless those details are cited directly in the source evidence above.