Best Chinese AI Company end of July?
Alibaba looks like the likeliest winner because Qwen is currently the highest-ranked Chinese model on Arena, but the lead is not so large that I would treat it as locked in.
Analysis
The latest Arena leaderboard snapshot I found is dated Jul 13, and it places Alibaba’s qwen3.7-max-preview at rank 16 overall with a 1475±10 score. The nearest Chinese peers on the same board are Z.ai’s glm-5.1 at rank 25, Baidu’s ernie-5.1 at rank 29, Xiaomi’s mimo-v2.5-pro at rank 32, Moonshot’s kimi-k2.6 at rank 37, and Tencent’s Hunyuan models much farther down, so Alibaba has a genuine lead among Chinese companies right now.
Alibaba also has the strongest visible release cadence among the Chinese contenders. Alibaba Cloud’s model docs were updated in early July, and Alibaba’s own chairman/CEO letter says the company released three Qwen updates in the past three months; those docs also describe qwen3.7-max as the latest flagship, built for agentic coding, complex reasoning, and long-horizon task execution. That matters because the market resolves on the July 31 Arena check, so a late Alibaba refresh is the most straightforward way to preserve first place.
The main reason not to make this closer to certain is that the Chinese model race is still moving fast. Tencent formally released Hy3 on July 6, DeepSeek is reported to be developing its own inference chip, and Reuters reported that Chinese authorities have discussed possible restrictions on overseas access to the country’s most advanced models, including unreleased ones; together, that means a surprise late-July preview from a rival is still plausible, even if no rival has yet displaced Qwen on the board. The market is already pricing a very high yes probability, but I think some extra caution is warranted because leaderboard positions near the top can shift quickly on a single strong release.
Arguments
For
- Arguments for Yes: Alibaba is already the highest-ranked Chinese company on the current Arena board, so it starts from first place.
- Arguments for Yes: Alibaba is still actively updating Qwen in early July, which raises the odds that it can defend the lead with another refresh before month-end.
- Arguments for Yes: The closest Chinese challengers are still several ranks behind, so they need a meaningful jump rather than a small nudge to overtake Alibaba.
- Arguments for Yes: Alibaba’s latest flagship is already positioned for the exact agentic and reasoning workloads that tend to score well in Arena-style comparisons.
Against
- Arguments against Yes: A single strong late-July preview from Z.ai, DeepSeek, or Tencent could leapfrog Qwen if the release is well tuned for Arena preferences.
- Arguments against Yes: qwen3.7-max-preview is still a preview listing, so it may not be as stable as a fully matured final release.
- Arguments against Yes: The top Chinese models are clustered enough that small Elo shifts can change the ranking order quickly.
- Arguments against Yes: Regulatory or access changes in China could distort when and how new frontier models appear on public leaderboards.
Key drivers
- Alibaba’s current lead on the Arena leaderboard is the single biggest factor supporting a Yes outcome.
- Alibaba’s ongoing Qwen update cadence gives it the best chance to defend the lead before the July 31 check.
- Competitors are active, but none have yet shown a public model that is close enough to dislodge Qwen immediately.
- Arena rankings can move on fresh previews, so the final two weeks are the key window for a challenger surprise.
Risk factors
- A late July model launch from a Chinese rival could overtake Qwen before the July 31 check.
- If Alibaba pauses updates or the preview does not improve, the current lead could erode quickly.
- Arena score compression near the top means a modest improvement from a rival may be enough to change the Chinese ranking order.
- Policy or access changes around Chinese frontier models could make the public ranking path less predictable than usual.
Scenarios
Best case
Alibaba ships a stronger Qwen snapshot or the current Qwen preview holds its edge, and no rival Chinese model gets a breakout Arena jump before July 31. In that case Alibaba finishes first among Chinese companies with some breathing room.
Most likely
Alibaba stays first among Chinese companies through the July 31 check, but the gap is narrow enough that the market will keep watching for a late preview update from a rival. My base case is Yes, with the main uncertainty coming from a surprise launch rather than from Qwen’s current standing.
Worst case
One of the faster-moving Chinese labs, most plausibly Z.ai, DeepSeek, or Tencent, releases a materially stronger model in late July and pushes above Qwen on the leaderboard. Alibaba would then lose the top Chinese spot even if its model remains competitive overall.
More from this day
- PoliticsKalshi1y
2026: Trump's dream year?
AI72%MKT6%Edge+66Hidden GemGiven that the three policy pillars of the bull case (tax cuts/deregulation, likely Fed rate cuts, and a weakening of regulatory restraints via the Supreme Court) are already in place, the outcome now primarily hinges on the 2026 midterms and a small residual macro/geo risk; I assess a substantially higher chance of the bull case occurring than the market does.
- PoliticsKalshi2y
Which agencies will Trump eliminate?
AI98%MKT36%Edge+62Hidden GemBased on clear executive actions, formal notifications, mass staff removals, and program cancellations in 2025, I assess with very high confidence that USAID was eliminated as an independent agency during Trump's term.
- techPolymarketEnded
Next Claude Opus released by...?
AI41%MKT97%Edge-56HypedThe market is pricing in a near-certain August Opus launch, but Anthropic's public materials still show Opus 4.8 as the current Opus model and there has been no official new Opus announcement in the past week. I think the true probability is much lower than the market's 97.55% implied price.