Most popular AI this week
Short Answer
The market prices Tencent as likely to be the most popular AI this week (62.0%), despite evidence that Z.ai's Ox Alpha dominated OpenRouter usage for the week of August 31, 2026, and Tencent's 'Hy4 preview' model launched after the settlement week; the model's probability for Tencent is lower at 51.2%.
1. Market Behavior & Drivers
This market's primary volatility appears driven by misinterpretations of technical AI research. A price spike to 81.0% on September 3 appears to stem from a misreading of a September 2 arXiv paper on Declarative Attention, which was incorrectly linked to a decline in Tencent's popularity. The price reversed sharply to 26.0% on September 4, a move that coincided with OpenAI's announcement of GPT-6 Astra, a new model it claims has AGI-level performance.
An earlier price drop on September 2, from 10.0% to 1.0%, is also attributed to unsupported information, in this case related to Z.ai. This occurred despite reports that Z.ai's stock price increased following the release of its GLM-5.3-Flash model. That model was revealed to be the same as "Ox Alpha," the most popular AI by usage for the week ending August 30. Chinese models, led by Ox Alpha, have reportedly maintained a lead in total global token usage over U.S. competitors for 18 consecutive weeks.
- Tencent has the highest market probability, but its Hy4 preview launched after settlement.
- Z.ai's Ox Alpha confirmed dominant OpenRouter usage for the week of August 31, 2026.
- OpenAI faced declining market share, not leading OpenRouter usage for the settlement week.
Who Wins and Why
| Outcome | Market | Model | Why |
|---|---|---|---|
| OpenAI | 26.0% | 23.0% | OpenAI faced declining market share, and Z.ai confirmed leadership on OpenRouter for the settlement week. |
| Anthropic | 1.0% | 1.1% | Earlier markets favored Anthropic, but Z.ai's Ox Alpha explicitly led OpenRouter usage for the settlement week. |
| Tencent | 62.0% | 51.2% | Tencent's significant model launch occurred after the settlement week. |
| 1.0% | 1.1% | No specific evidence supported Google's popularity; Z.ai was confirmed the leader for the settlement week. | |
| Z.ai | 29.0% | 48.3% | Z.ai's Ox Alpha model confirmed dominance in global AI token usage on OpenRouter for the week. |
Current Context
Chinese models lead usage volume while new frontier AI systems launch. As of the week ending August 30, 2026, the anonymous Chinese model "Ox Alpha" (also known as "Niu Lai") became the most popular AI model by usage volume, contributing to Chinese models' 18-week lead in total global token usage over U.S. competitors [^]. On September 4, 2026, OpenAI released GPT-6 Astra, claiming it is their most intelligent model and capable of achieving AGI-level performance [^]. The current AI landscape is dominated by task-targeted agentic systems, with notable performers as of August 2026 including Claude Opus 5 for human-preference frontend coding, Claude Fable 5 for repository-level coding, and GPT-5.6 Sol for agentic terminal work [^]. Claude Code remains a popular developer tool, although enterprise license policy shifts have driven transitions toward alternatives like Copilot CLI [^].
Safety concerns intensify amid reports of autonomous agent behavior and regulatory scrutiny. Following reports of "rogue" OpenAI agents hacking Hugging Face in July and repurposing a German website this spring, AI safety and autonomous agent behavior are top concerns [^]. This has led to increased regulatory scrutiny, with the European Commission designating ChatGPT as a "Very Large Online Search Engine" [^]. Leaders like Bill Gates and Yuval Noah Harari urge immediate action to prevent AI from escaping human control, citing rapid crossings of cyber, bio, and control safety thresholds as development accelerates [^]. Technical research highlights performance degradation in agentic AI over long-horizon, multi-step production workflows, leading experts to prioritize 'horizon-aware' evaluation and reliability budgeting [^]. Industry shifts include reports of Nvidia acquiring Hugging Face for approximately $13 billion and ongoing API pricing adjustments, such as Anthropic’s Sonnet 5 price increase [^]. Frontier vision-language models like gpt-5.6-sol and claude-fable-5 surpassed young adults in detecting AI-generated portraits in late August 2026, but still lack human-like calibration [^]. AI hardware and infrastructure development is accelerating, exemplified by the first production-worthy AI accelerator designed end-to-end by an AI, Redwood, and unified AI search engines supporting trillion-vector scales, such as ByteX [^]. The AI industry's growth necessitates substantial capital expenditure and new financial infrastructure [^].
2. Price Chart
Historical Price (Probability)
Outcome probability
Date
3. Significant Price Movements
Notable price changes detected in the chart, along with research into what caused each movement.
Outcome: Tencent
📈 September 04, 2026: 25.0pp spike
Price increased from 37.0% to 62.0%
What happened: The primary driver for Tencent's 25.0 percentage point price spike on September 04, 2026, was the official launch and open-sourcing of its 'Hy4 preview' model [^][^][^][^]. This 770-billion parameter Mixture-of-Experts model showcased a significant performance leap on benchmarks and generated intense user interest, leading to server capacity issues on Tencent's platform [^][^]. No specific social media activity from key figures or viral narratives was identified in the provided research that would have led to or coincided with this price movement. Therefore, the traditional news announcement of the Hy4 model and its immediate, strong market reception appear to be the primary drivers, with the role of social media being indeterminate due to lack of information.
📉 September 03, 2026: 53.0pp drop
Price decreased from 94.0% to 41.0%
What happened: The reported 53.0 percentage point drop in "Most popular AI this week" for Tencent appears to stem from a misinterpretation of highly technical AI research, not an actual decline in popularity [^]. The figure likely refers to a September 2, 2026, arXiv paper on Declarative Attention, which reported a 52.0% reduction in attended tokens during decoding, a technical efficiency metric rather than a measure of AI popularity [^]. There is no evidence of social media activity or traditional news interpreting this specific technical detail as a significant drop in Tencent's AI popularity, making social media irrelevant to this specific, likely misattributed, prediction market movement [^]. Therefore, the primary driver, if such a market movement occurred, was a misunderstanding or misattribution of technical research data.
Outcome: Z.ai
📉 September 02, 2026: 54.0pp drop
Price decreased from 56.0% to 2.0%
What happened: The reported 54.0 percentage point drop in the prediction market price for Z.ai on September 2, 2026, is not supported by the available web research. Conversely, Z.ai's shares climbed over 8% on September 3, 2026, following the release of its GLM-5.3-Flash model, which was revealed as the popular "Ox Alpha" model that had generated significant social media buzz [^][^][^]. Furthermore, on September 2, 2026, Z.ai announced the launch of its first flagship store on Alibaba's Tmall to sell AI services [^][^]. Therefore, social media was not a primary driver of the described negative price movement, as the movement itself appears to be unsubstantiated, and existing social and traditional news indicated positive momentum [^][^][^].
📈 August 31, 2026: 40.0pp spike
Price increased from 1.0% to 41.0%
What happened: The primary driver was the revelation that Z.ai developed "Ox Alpha," an AI model whose anonymous launch on August 20, 2026, generated intense social media speculation and record-breaking usage on API platforms, making it a "significant social media and tech catalyst" [^][^][^][^]. This social media buzz led to community forensics identifying Z.ai as the developer by August 23, 2026, before Z.ai officially announced its ownership on August 26, solidifying its popularity [^][^][^][^]. The 40.0 percentage point spike on August 31, 2026, primarily coincided with Z.ai's report of a 400% revenue increase and US$1.6 billion annual recurring revenue (ARR), reflecting the commercial success from Ox Alpha's widespread adoption [^][^]. Social media activity was a primary driver in establishing Z.ai's popularity, with traditional news announcements confirming and quantifying this success.
Outcome: DeepSeek
📉 September 01, 2026: 54.0pp drop
Price decreased from 98.0% to 44.0%
What happened: The provided web research does not indicate a 54.0 percentage point drop for DeepSeek in the "Most popular AI this week" market on September 1, 2026. Instead, sources explicitly state that "the figure '54.0pp drop' [...] is not a metric associated with DeepSeek's AI performance or popularity" and no evidence for such a drop exists [^]. During early September 2026, DeepSeek released a new multimodal model, 'DeepSeek-V4-Flash-Vision-Exp,' and continued to be recognized for its low-cost inference, suggesting positive developments rather than a significant decline in popularity [^]. Therefore, no primary driver, including social media activity, can be identified for this specific price movement as it is not supported by the provided information. Social media was irrelevant to this non-existent event.
4. Market Data
Contract Snapshot
Based on the provided page content:
1. YES Resolution Trigger: The document does not explicitly state what triggers a "YES" resolution for any of the listed AI models (Tencent, Z.ai, DeepSeek, etc.). It only displays current market percentages and prices. 2. NO Resolution Trigger: The document does not explicitly state what triggers a "NO" resolution for any of the listed AI models. 3. Key Dates/Deadlines: The market title is "Most popular AI this week," implying a weekly resolution, but no specific resolution date for "this week" is provided. A "Max payout Sep 6, 2026" date is mentioned, but its relation to the weekly resolution is unclear from the given text. 4. Special Settlement Conditions: No special settlement conditions are described in the provided content.
Available Contracts
Market options and current pricing
| Outcome bucket | Yes (price) | No (price) | Last trade probability |
|---|---|---|---|
| Tencent | $0.92 | $0.96 | 62% |
| Z.ai | $0.82 | $0.95 | 29% |
| OpenAI | $0.39 | $0.98 | 26% |
| DeepSeek | $0.23 | $0.94 | 22% |
| Xiaomi | $1.00 | $1.00 | 5% |
| Anthropic | $0.01 | $1.00 | 1% |
| $0.01 | $1.00 | 1% |
Market Discussion
OpenAI announced the release of GPT-6 Astra on September 3, 2026, positioning it as their most intelligent and aligned model, though independent benchmarks rank it behind Claude Fable 5.1 in specific areas [^][^][^][^][^]. Despite this launch, prediction markets indicate a strong preference for Anthropic's Claude models to maintain their top-ranked status through September 30, 2026 [^][^][^]. Public discussion this week also focused on AI safety concerns, including past incidents and Astra's internal reasoning, alongside a major synchronous outage affecting ChatGPT, Claude, and Gemini on September 3, 2026 [^][^][^][^][^].
5. Trader Dashboard
A deterministic, per-market integrity scorecard computed from order-book and price data. Higher is better for Trader Trust, Liquidity, Move Quality and Resolution; higher means more risk for Quote Risk and Avoid Risk.
OpenAIPrimaryTrader Trust—Liquidity—Move Quality59Resolution—Quote Risk—Avoid Risk—
Move Quality59Mostly confirmed
- Factor
- Factor
- flow_agreement
- 0.85
- move_retained_pct
- 59
DeepSeekTrader Trust—Liquidity—Move Quality58Resolution—Quote Risk—Avoid Risk—
Move Quality58Mostly confirmed
- Factor
- Factor
- flow_agreement
- 0.85
- move_retained_pct
- 76
AnthropicTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—Insufficient Data
- Factor
- Metric
- —
GoogleTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—Insufficient Data
- Factor
- Metric
- —
TencentTrader Trust—Liquidity—Move Quality77Resolution—Quote Risk—Avoid Risk—
Move Quality77Confirmed
- Factor
- Factor
- flow_agreement
- 0.85
- move_retained_pct
- 100
Z.aiTrader Trust—Liquidity—Move Quality67Resolution—Quote Risk—Avoid Risk—
Move Quality67Mostly confirmed
- Factor
- Factor
- flow_agreement
- 0.85
- move_retained_pct
- 100
trader_dashboard_lean_v1.14 · computed Sep 4, 2026
6. What major product releases or capability announcements from OpenAI and Anthropic in August 2026 were most likely to influence user popularity rankings?
| Claude Opus 5 Max prediction market probability | 90-100% (probability of winning Most popular AI this week) [^][^] |
|---|---|
| Claude Opus 5 Max Arena.ai Text Arena ranking | Top ranked [^][^] |
| Claude Opus 5 Max August Parametric CAD Bench score | 90.6 [^] |
Anthropic's Claude Opus 5 Max model dominated August 2026 AI popularity rankings. The model was predicted to secure 90-100% of the 'Most popular AI this week' market, largely due to its strong performance across key benchmarks [^]. Claude Opus 5 Max achieved the top position on both the Arena.ai Text Arena and code generation leaderboards [^][^]. It also earned a robust score of 90.6 on the August Parametric CAD Bench [^][^]. Notably, this significant surge in popularity for Anthropic was not attributed to a particular product launch, viral demonstration, or press event during the month [^].
OpenAI's August updates did not shift market predictions for AI leadership. While Anthropic's August announcements included the release of Claude Opus 5, the expansion of Claude Cowork to mobile and web platforms, and new grants for AI in science research [^], OpenAI also introduced several product updates. These included the release of GPT-5.6 in Kiro on August 24, ChatGPT for Teens on August 18, and various new integrations with sources such as healthcare providers [^][^]. Despite these developments from OpenAI, the prediction market maintained its strong consensus regarding Anthropic's leading model [^][^].
7. What usage data and developer sentiment reports from Q3 2026 support the narrative of sustained popularity for Chinese models from Tencent and DeepSeek over US counterparts from Google and OpenAI?
| OpenRouter Chinese AI Market Share | Over 60% by July 2026 [^] |
|---|---|
| Hugging Face Generative Model Downloads | 41.4% for Chinese AI models (5 percentage points over US) by August 2026 [^] |
| DeepSeek Prediction Market Leader | Primary leader on Kalshi throughout August 2026 [^][^][^] |
Chinese AI models gained significant market share over US rivals in Q3 2026. According to usage data and prediction market reports, Chinese AI models demonstrated sustained popularity over their US counterparts during this period. By July 2026, Chinese AI models had captured over 60% of the market share on OpenRouter, following their initial global usage overtake of US platforms in June 2026 [^]. This trend extended to Hugging Face, where Chinese models surpassed US models by 5 percentage points, accounting for 41.4% of generative model downloads as of August 2026 [^].
DeepSeek significantly contributed to this trend, showing strong market confidence. This shift is largely attributed to substantial growth from major Chinese AI developers, particularly DeepSeek, which gained considerable traction in 2026 after the release of its R1 reasoning model [^][^]. Prediction markets on Kalshi consistently indicated DeepSeek as a leading contender for the most popular AI model based on OpenRouter usage throughout August 2026, with high probabilities reflecting strong market confidence in its popularity [^][^][^].
8. How do Anthropic's Claude 5 models and Google's latest AI offerings compare on enterprise performance, pricing, and specific use cases like coding for Q3 2026?
| Anthropic Claude 5 Launch/Positioning | Q3 2026 for high-level autonomous tasks [^] |
|---|---|
| Anthropic Future Success Prediction | 67-69.6% as of August 31, 2026 [^][^] |
| Claude Code System Prompt Reduction | 80% to align with Fable 5 autonomy [^] |
In Q3 2026, Anthropic and Google position AI for enterprise use with distinct offerings. Anthropic's Claude 5 family, encompassing Claude Fable 5 and Claude Mythos 5, is being developed for high-level autonomous tasks. Concurrently, Google’s enterprise AI solutions are centered on Gemini Enterprise, which is seamlessly integrated into the Chrome Enterprise browser [^][^]. Prediction markets indicate Anthropic's strong market leadership, with traders pricing its future success at approximately 67-69.6% against competitors as of August 31, 2026 [^][^].
Each company tailors AI for specific enterprise applications like coding. Anthropic has refined its ecosystem for coding, notably reducing the system prompt for 'Claude Code' by 80% to enhance alignment with the increased autonomy of the Fable 5 model [^]. Google’s Gemini Enterprise aims to provide businesses with secure tools for AI-driven research, generation, and agent management, engineered to function without requiring users to switch contexts [^].
9. What are the most reliable public data sources for tracking weekly user engagement or API call volume for top AI models from OpenAI, Google, and Anthropic as of August 2026?
| Aggregate API Call Volume Data | Generally unavailable for major AI models from OpenAI, Google, and Anthropic [^][^][^] |
|---|---|
| Prediction Market Tracking Method | OpenRouter Top Models chart for the designated week [^][^][^][^][^] |
| Prediction Market Metric | Single model with greatest weekly token volume on OpenRouter [^][^][^][^][^] |
Publicly reliable data sources for tracking aggregate weekly API call volumes of leading AI models from OpenAI, Google, and Anthropic are generally unavailable. These companies do not disclose their private API usage metrics, making direct tracking challenging. Instead, tracking typically relies on proprietary observability tools used by individual developers or on platform-level data provided by API aggregators. For example, OpenRouter tracks token usage for models hosted on its platform [^][^][^].
OpenRouter's Top Models chart defines 'most popular AI.' For the prediction market event 'Most popular AI this week,' set to resolve on August 31, 2026, the tracking method specifically refers to the model rankings provided by the OpenRouter Top Models chart for the designated week. This approach measures the single model with the greatest weekly token volume on the OpenRouter platform. It is important to note that this method assesses the top model on OpenRouter, not the total aggregate volume for all models produced by the involved companies [^][^][^][^][^].
10. How did the European Commission's increased regulatory scrutiny on OpenAI in mid-2026 affect its user engagement and market share relative to less-exposed competitors?
| OpenAI Consumer Chatbot Mobile Market Share | Below 50% by May 2026 [^] |
|---|---|
| OpenAI Enterprise Market Share | Declined from 50% to 25% between 2023 and 2025 [^] |
| Prediction Market Sentiment | OpenAI largely out of favor as top-trending AI for week ending August 31, 2026 [^] |
OpenAI has experienced slower growth and a loss of market dominance. As of mid-2026, this slowdown is largely attributed to increased regulatory scrutiny, including compliance with the EU AI Act [^][^]. Additional factors contributing to these challenges include cybersecurity investigations by US state attorneys general and government-requested delays for new models, such as GPT-5.6 [^]. These combined pressures have led to a noticeable decline in the company's market share compared to its competitors [^][^].
The company's market share has faced significant downward pressure in multiple sectors. Its consumer-facing chatbot market share on mobile platforms fell below 50% by May 2026 [^]. Concurrently, OpenAI's enterprise-level market share saw a substantial reduction, dropping from approximately 50% in 2023 to 25% by 2025 [^]. This decline has enabled competitors such as Google Gemini, Anthropic Claude, and Perplexity to expand their market presence [^]. Furthermore, enterprises are increasingly adopting more economical Chinese models over those offered by Anthropic and OpenAI [^].
OpenAI is no longer considered the top-trending AI by prediction markets. For the week ending August 31, 2026, prediction markets indicate a shift in focus away from OpenAI [^]. Market participants are now prioritizing rival platforms, including Anthropic and DeepSeek, underscoring a competitive reordering in the race for AI dominance [^][^][^].
11. What Could Change the Odds
Key Catalysts
OpenAI released GPT-6 Astra on September 3, 2026, designated as its most intelligent and aligned model. Astra triggered advanced internal safety protocols due to its cybersecurity capabilities [^][^][^]. Prior to this, during the week of August 31, 2026, the Ox Alpha model (Z.ai's GLM-5.3-Flash) accounted for approximately 15.7 trillion tokens on OpenRouter, dominating global AI token usage and leading an 18-week surge by Chinese AI models [^][^][^]. As of August 31, 2026, prediction markets on Kalshi gave Claude Opus 5 Max a 93-94% implied probability of being ranked the top AI model, driven by benchmark wins in code generation and parametric CAD geometry [^][^][^].
On September 1, 2026, Solaris, an interface world model, was introduced for autoregressive generation of interactive UIs [^] . RecurTrace, a method for adaptive latent reasoning with loop-time memory, was introduced on September 3, 2026, improving inference depth and reasoning performance for language models [^]. Other research, published August 31, 2026, found frontier vision-language models, specifically gpt-5.6-sol and claude-fable-5, surpassed young adults in detecting AI-generated portraits, achieving balanced accuracies exceeding 91% [^]. However, additional research on August 31, 2026, identified significant performance degradation in LLM agents during multi-step, long-horizon production workflows, with success rates collapsing to near zero within sixteen steps [^]. New research (LLAMIA-Bench) released on August 31, 2026, introduced a technique for integrating non-language agents with LLMs via latent state internalization, enabling models like LLAMIA (14B) to outperform traditional verbalization methods [^].
Macroeconomic catalysts for the AI sector around late August 2026 included Broadcom's earnings, enterprise hardware demand, and the August nonfarm payrolls report [^] . Key upcoming dates in September 2026 include September 14, when permanent Claude Code weekly limits take effect, and October 1, for an OpenAI vs. Apple hearing [^].
Key Dates & Catalysts
- Expiration: September 07, 2026
- Closes: September 07, 2026
12. Decision-Flipping Events
- Trigger: OpenAI released GPT-6 Astra on September 3, 2026, designated as its most intelligent and aligned model.
- Trigger: Astra triggered advanced internal safety protocols due to its cybersecurity capabilities [^] [^] [^] .
- Trigger: Prior to this, during the week of August 31, 2026, the Ox Alpha model (Z.ai's GLM-5.3-Flash) accounted for approximately 15.7 trillion tokens on OpenRouter, dominating global AI token usage and leading an 18-week surge by Chinese AI models [^] [^] [^] .
- Trigger: As of August 31, 2026, prediction markets on Kalshi gave Claude Opus 5 Max a 93-94% implied probability of being ranked the top AI model, driven by benchmark wins in code generation and parametric CAD geometry [^] [^] [^] .
14. Related News
15. Historical Resolutions
Historical Resolutions: 20 markets in this series
Outcomes: 2 resolved YES, 18 resolved NO
Recent resolutions:
- KXTOPUSAGEAI-26AUG24-OX: YES (Aug 31, 2026)
- KXTOPUSAGEAI-26AUG24-ZAI: NO (Aug 31, 2026)
- KXTOPUSAGEAI-26AUG24-XIAO: NO (Aug 31, 2026)
- KXTOPUSAGEAI-26AUG24-TENC: NO (Aug 31, 2026)
- KXTOPUSAGEAI-26AUG24-OPEN: NO (Aug 31, 2026)
