As of the latest public Arena snapshot (Sep 25, 2026) Xiaomi’s MiMo‑V2.6‑Pro is the highest-ranked primarily Chinese model (Rank 9, 1491±9) while Alibaba’s best Qwen variant (qwen3.8‑max) sits lower (Rank 17, 1480±5). Alibaba has recently released and upgraded Qwen3.8 variants and made them available on cloud/model portals, so a late surge (new snapshot, added variant, or concentrated Arena evaluation) between Sep 26 and the resolution time could allow Alibaba to become the top Chinese entry by Sep 30, 2026. However, Xiaomi’s very recent open-source MiMo‑V2.6 release and the presence of other competitive Chinese models make the intra‑Chinese ranking tightly contested and sensitive to short-term voting dynamics on Arena. The market’s resolution is objectively checkable at 2026-09-30 12:00 PM ET per the specified Arena configuration, but the outcome depends primarily on whether Alibaba’s Qwen variants overtake Xiaomi (or other Chinese competitors) in Arena pairwise votes during the remaining window and on Arena’s availability and exact configuration at the check time.
a9-predict-0.3.0 · c41eb3cfa9c5560df41462cebde49710c148942be7622a3798bce270a62622c0Tracked question
Will Alibaba have the best Chinese AI model at the end of September 2026?
A9 gives this a 11% chance of happening. That's about the same as the betting market (9%). A9 researched this on the web before forecasting.
Jev on its own said 17%
Why A9 thinks this
The evidence, separated.
As of the latest public Arena snapshot (Sep 25, 2026) Xiaomi’s MiMo‑V2.6‑Pro is the highest-ranked primarily Chinese model (Rank 9, 1491±9) while Alibaba’s best Qwen variant (qwen3.8‑max) sits lower (Rank 17, 1480±5). Alibaba has recently released and upgraded Qwen3.8 variants and made them available on cloud/model portals, so a late surge (new snapshot, added variant, or concentrated Arena evaluation) between Sep 26 and the resolution time could allow Alibaba to become the top Chinese entry by Sep 30, 2026. However, Xiaomi’s very recent open-source MiMo‑V2.6 release and the presence of other competitive Chinese models make the intra‑Chinese ranking tightly contested and sensitive to short-term voting dynamics on Arena. The market’s resolution is objectively checkable at 2026-09-30 12:00 PM ET per the specified Arena configuration, but the outcome depends primarily on whether Alibaba’s Qwen variants overtake Xiaomi (or other Chinese competitors) in Arena pairwise votes during the remaining window and on Arena’s availability and exact configuration at the check time.
For
Arena.ai Text Arena (Overall) snapshot on Sep 25, 2026 shows multiple Alibaba Qwen variants present on the leaderboard (qwen3.8-max at Rank 17, qwen3.7-max at Rank 21, qwen3.5 variants at Rank 28/43), demonstrating Alibaba’s active presence and relatively high placements among lab entries as of the latest snapshot.
Alibaba publicly released Qwen3.8-Max (flagship of Qwen-3 family) earlier in 2026 and published upgraded snapshots (e.g., Qwen3.8-Max-0902) and platform availability (Model Studio / AI Gateway), indicating the company has recently shipped higher-capability variants that could improve Arena performance before Sep 30, 2026.
Alibaba has a rapid release and upgrade cadence (announcements and model-snapshot upgrades documented across Alibaba Cloud pages and press), meaning near-term model improvements or additional Qwen variants could be added to Arena between Sep 26 and Sep 30.
Arena’s methodology is pairwise-vote / Elo-style and allows newly added or upgraded public models to be incorporated and scored; Alibaba’s recent publication of open weights / snapshots increases the chance community evaluators will sample and re-evaluate Qwen variants on Arena promptly.
Against
Arena.ai Text Arena (Overall) snapshot on Sep 25, 2026 shows Xiaomi’s MiMo-V2.6-Pro at Rank 9 with score 1491±9 — currently the highest-ranked primarily Chinese model on that leaderboard, ahead of Alibaba’s top Qwen entry (qwen3.8-max at Rank 17, 1480±5).
Xiaomi publicly released and open-sourced the MiMo-V2.6 series on Sep 21–22, 2026 (Pro and Flash), and early coverage and community uptake (open weights + Hugging Face / platform availability) make it likely Xiaomi’s models will retain or increase Arena visibility and voting before Sep 30.
Other primarily Chinese labs (Z.ai’s GLM-5.3, Baidu Ernie-5.1, DeepSeek, Moonshot/Kimi K3, ByteDance) already occupy leaderboard positions close to or above some Alibaba entries, making intra‑Chinese competition tight and ranking sensitive to a small number of additional votes or runs.
Arena leaderboards are dynamic and driven by human pairwise voting and methodology updates; short-term vote swings, sampling-frequency effects for newly added models, or confidence‑interval overlap can change rank order rapidly, making current gaps surmountable and also easily eroded.
Unknown
How Arena voting and sampling for the Text Arena will change between Sep 26 and the resolution check on Sep 30, 2026 (which models receive evaluation traffic, whether Alibaba publishes any new model snapshot or variant that is added to Arena before the check, and whether Arena is online and applying the same leaderboard configuration at 2026-09-30 12:00 PM ET).
What would change A9's mind
- Alibaba releases (or Arena ingests) a materially upgraded Qwen snapshot or new Qwen variant between Sep 26 and Sep 30 and that variant receives enough pairwise evaluations on Arena to push its score above Xiaomi’s top Chinese model.
- Xiaomi’s MiMo-V2.6 (or another Chinese model) receives a surge of positive pairwise votes or sampling bias that raises its Arena score further, making it harder for Alibaba to overtake by Sep 30.
- Arena changes its scoring configuration, sampling, or 'Style Control' / 'Adjustments' defaults (the resolution specifies 'style control off' and 'Adjustments: None' — any deviation or site misconfiguration at check time could affect rank ordering).
- Arena is unavailable at check time (per the market rules, if Arena is offline the market remains open until the leaderboard returns, or resolves to 'Other' if permanently unavailable).
- A tie on rank occurs among Chinese-company models and Arena’s underlying granular scores (or the market’s specified tiebreak rules) are required to break the tie — small numeric differences in unrounded scores could flip the winner.
Probability history