Best Chinese AI Company end of July?
Alibaba remains the strongest bet to finish July with the top Chinese model on LMArena. I still give a small haircut to the market price because the lead is real but not immune to a late-month challenger or score reshuffle.
Analysis
As of the July 10 Text Arena leaderboard, Alibaba’s qwen3.7-max-preview is the highest-ranked Chinese model at rank 17 overall, ahead of Z.ai’s glm-5.1 at 25, Baidu’s ernie-5.1 at 30, Xiaomi’s mimo-v2.5-pro at 31, Moonshot’s kimi-k2.6 at 37, and Alibaba’s own qwen3.6-max-preview at 38. That is a meaningful cushion inside the Chinese cohort, even though the scores near the top are still close enough that the order is not completely locked in.
Alibaba’s July 8 cloud documentation says qwen3.6-max-preview is scheduled for deprecation on October 10, 2026 and will be replaced by qwen3.7-max, which signals that Qwen3.7 is the company’s current mainline path rather than a stopgap. Alibaba’s May 20 launch announcement also framed Qwen3.7-Max as its flagship model for agentic workloads, so the company has both product momentum and a clear incentive to keep the family competitive through month-end.
The main reason not to call this certain is that qwen3.7-max-preview is still marked preliminary, and its confidence band is wide enough that modest vote accumulation could compress the gap with nearby rivals. The broader Chinese AI field is also still active: Moonshot announced Kimi K2.7 Code on June 16 and Z.ai launched GLM-5.2 on June 16, so the biggest threat is a surprise late-July release or a fast follow-up from one of those labs rather than a slow drift in the current board. Given that the market already prices Yes at 94.7%, I agree with the direction but keep a little extra room for a late move.
Arguments
For
- Alibaba currently holds the highest-ranked Chinese model on the latest overall Text Arena leaderboard, so it enters the final stretch in first place.
- Qwen3.7-Max is the active successor to Alibaba’s older Qwen3.6-Max line, which suggests ongoing support and attention from the company.
- The next Chinese competitors are still several ranks behind on the current board, giving Alibaba a practical buffer against ordinary vote drift.
Against
- qwen3.7-max-preview is still preliminary, so the leaderboard can still move as more votes are collected.
- The score gap over nearby Chinese rivals is not enormous, which leaves room for a rank flip if their confidence intervals tighten in the other direction.
- Moonshot and Z.ai both continue to ship frontier models, so a late-July release could change the Chinese order quickly.
Key drivers
- Current leader status is the single strongest factor in Alibaba’s favor.
- Alibaba’s recent deprecation notice confirms that qwen3.7-max is the company’s active replacement path.
- No rival Chinese lab has shown a clearly superior general text model on the latest Arena board.
- Alibaba’s ongoing agentic model push makes another quality refresh before month-end plausible.
Risk factors
- A preliminary model with a wide score interval is inherently vulnerable to short-term rank changes.
- A surprise release from Moonshot, Z.ai, or another Chinese lab could overtake Qwen before July 31.
- Leaderboard ranks can tighten quickly when new votes arrive, especially for models with fewer observations.
- If Alibaba’s current model stops improving while rivals keep iterating, the current lead could erode.
Scenarios
Best case
Alibaba keeps the top Chinese spot through the July 31 check, and qwen3.7-max-preview either stays ahead or improves slightly as more votes come in.
Most likely
Alibaba remains first among Chinese companies at the end of July, but the margin may narrow if the leaderboard gets another wave of votes or a competitor refreshes its flagship model.
Worst case
A new Chinese general-purpose release or a score-driven reshuffle pushes a rival above qwen3.7-max-preview before month-end, ending Alibaba’s lead.
More from this day
- HealthKalshi2y
What will the average number of measles cases be during Trump's term?
AI72%MKT28%Edge+44Hidden GemI assess a 72% independent probability that the average annual number of US measles cases across 2025–2028 will meet or exceed ~3,000 per year (i.e., the 'Yes' outcome). This reflects the strong acceleration in 2026, weak policy response, and high-risk pockets that make a sustained multi-year elevated average more likely than not.
- pop culturePolymarketEnded
What will happen before GTA VI?
AI9%MKT51%Edge-42HypedI think the chance of another pandemic before GTA VI is materially below the current market price. The most likely outcome is that today’s outbreaks remain serious but contained, with no event reaching true pandemic status before November 19, 2026.
- EconomicsKalshi2y
Will there be a Trump economic boom?
AI8%MKT50%Edge-42HypedIndependent assessment: very unlikely — I assign an 8% chance that any U.S. quarterly real GDP (annualized) will exceed 5% in Q1 2025–Q4 2028. The data and policy environment point strongly against a sustained or one‑off 5%+ quarter, though low‑probability shocks could still produce a spike.