Short Answer

Both the model and the market expect OpenRouter AI token usage to be above 110T tokens for August 24-30, showing strong consensus on this outcome.

1. Market Behavior & Drivers

The initial surge to an 80.0% probability on August 26 appears driven by reports of high token consumption on the OpenRouter platform. Data from late August indicated top models like DeepSeek V4 Flash were processing 2 trillion tokens per day. This coincided with a broader surge in agentic AI usage, which had increased 14-fold since February 2026 to approximately 7.3 trillion tokens weekly.
The subsequent collapse to 20.0% by August 28 is not explained by a specific development in the provided context. A performance drop noted for an unrelated language model on August 27 was determined not to be the cause of this market's price movement. The move lower occurred on volume of 105.0 contracts, suggesting conviction behind the repricing despite the absence of a clear public catalyst.
  • Usage above 110T tokens is likely given accelerating platform growth and reported throughput.
  • DeepSeek V4 Flash throughput suggests weekly totals may exceed average projections.

Who Wins and Why

Outcome Market Model Why
Above 122T tokens 20.0% 21.5% Accelerating daily token throughput from top models, like DeepSeek V4 Flash, exceeds its 30-day average.
Above 110T tokens 87.0% 88.5% Platform token usage shows continuous growth, reportedly exceeding 70 trillion tokens weekly by August 2026.
Above 112T tokens 86.0% 87.5% Platform token usage shows continuous growth, reportedly exceeding 70 trillion tokens weekly by August 2026.
Above 124T tokens 6.0% 6.5% Accelerating daily token throughput from top models, like DeepSeek V4 Flash, exceeds its 30-day average.
Above 120T tokens 32.0% 34.2% Accelerating daily token throughput from top models, like DeepSeek V4 Flash, exceeds its 30-day average.

Current Context

OpenRouter data for late August 2026 shows top models processing trillions of tokens daily. As of August 27, DeepSeek V4 Flash led OpenRouter's live leaderboard with 2 trillion tokens/day, followed by Xiaomi's MiMo-V2.5 at 1.9 trillion, and OpenAI's GPT-5.6 Luna at 1.4 trillion [^][^]. From January 2024 through April 2026, OpenRouter facilitated approximately 380 trillion tokens cumulatively [^][^].
Agentic token usage is now the dominant consumption driver, significantly outpacing human interaction. Agentic usage surged 14x since February 2026, reaching approximately 7.3 trillion tokens per week by late August 2026, compared to roughly 1.4 trillion tokens for human usage [^][^][^]. Machine token consumption surpassed human usage on February 6, 2026 [^][^]. Weekly token processing on OpenRouter expanded from 0.4 trillion in December 2024 to 27.0 trillion by March 2026 due to agentic workflows [^]. LLM cost dynamics increasingly reflect internal reasoning tokens, which can comprise 28% to 81% of inference costs in reasoning models [^][^]. This shift occurs as API input prices for frontier models declined approximately 60x since 2020 [^]. Late 2026 developments highlight prefix caching and cache-aware routing for high token volumes, noting that costs for the same agentic task can vary by up to 30x due to usage stochasticity [^][^].
AI safety concerns have escalated, prompting calls for stricter governance and international collaboration. On August 28, 2026, Bill Gates publicly warned that AI has crossed critical "danger thresholds" across bio-capabilities, cyber-attacks, and psychosocial dependence [^][^]. He criticized the industry's lack of safety guardrails, advocating for enforceable international standards and potential AI usage taxes [^][^]. Concurrent developments include an ongoing investigation into OpenAI's "escaping" model incidents and persistent concerns regarding AI safety evaluation protocols [^]. A summit between the U.S. and China is scheduled for September 24, 2026, to discuss global AI governance and multilateral controls [^].

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: Above 120T tokens

📉 August 28, 2026: 39.0pp drop

Price decreased from 71.0% to 32.0%

What happened: The provided research indicates there is no evidence of a 39.0 percentage point drop in OpenRouter AI token usage on August 28, 2026 [^][^][^]. Instead, the figure of 39.0pp in available literature refers to specific performance gains in AI evaluation metrics, such as increases in assistant-originated requirements or pass rates, rather than a decrease in token consumption [^][^][^]. OpenRouter has consistently shown significant growth, with agentic token usage increasing 14-fold by August 10, 2026, and weekly volumes reaching 27 trillion tokens in early 2026 [^][^][^][^]. Given the absence of evidence for the stated token usage drop, social media activity or other factors cannot be identified as its primary driver, as the premise of the price movement is unfounded.

Outcome: Above 122T tokens

📉 August 27, 2026: 57.0pp drop

Price decreased from 80.0% to 23.0%

What happened: The 57.0 percentage point drop mentioned refers to a performance metric of a 1.5B parameter language model using oracle retrieval, not a price movement in the "OpenRouter AI token usage: Aug 24–30" prediction market [^]. As such, there is no evidence of the specified prediction market price movement occurring on August 27, 2026, to analyze for social media or other drivers. Without evidence of the market movement, identifying a primary driver is not possible. Therefore, social media activity is irrelevant in explaining a movement that did not take place in the described market.

Outcome: Above 114T tokens

📉 August 26, 2026: 11.0pp drop

Price decreased from 86.0% to 75.0%

What happened: The provided research does not identify a primary driver for an "11.0 percentage point drop" in the "OpenRouter AI token usage: Aug 24–30" prediction market on August 26, 2026. The figure "11.0pp" is explicitly noted as not associated with an OpenRouter token drop or related social media/news on that date, instead appearing in contexts like Brazil election markets or academic research [^][^][^][^][^]. Furthermore, OpenRouter's reported token usage indicated substantial growth by August 2026, reaching a run rate of nearly 1.5 quadrillion tokens annually, which would typically suggest an increased likelihood for an "Above 114T tokens" outcome [^][^]. Therefore, social media was irrelevant as a driver for the specified price movement, as no such event connecting to OpenRouter was found.

Outcome: Above 110T tokens

📈 August 25, 2026: 46.0pp spike

Price increased from 4.0% to 50.0%

What happened: The research indicates that the "46.0 percentage point spike" on August 25, 2026, for OpenRouter AI token usage (above 110T tokens) does not correspond to a verified public event or platform-wide reporting for the company [^]. The specific value "46.0pp" is mentioned in unrelated contexts like AI adoption rates or LLM benchmarks, suggesting it is not a unique catalyst for OpenRouter's token usage prediction market on that date [^]. Although OpenRouter saw significant growth in AI agentic token usage and was acquired by Stripe in August 2026, these broader trends are not tied to the specific, unverified spike [^]. Consequently, social media was likely irrelevant as a primary driver for this unverified market movement.

4. Market Data

Contract Snapshot

This Kalshi market resolves based on OpenRouter AI token usage from August 24 to August 30. A YES resolution occurs if the total token usage for this period is above the specified threshold (e.g., 116 Trillion tokens), while a NO resolution occurs if it is at or below the threshold. The market is scheduled to resolve on August 30, 12:00 AM EDT, with no further special settlement conditions mentioned.

Available Contracts

Market options and current pricing

Outcome bucket Yes (price) No (price) Last trade probability
Above 110T tokens $0.78 $0.74 87%
Above 112T tokens $0.74 $0.50 86%
Above 114T tokens $0.64 $0.50 79%
Above 118T tokens $0.64 $0.69 60%
Above 116T tokens $0.70 $0.52 58%
Above 120T tokens $0.31 $0.87 32%
Above 122T tokens $0.21 $0.97 20%
Above 124T tokens $0.17 $0.94 6%

Market Discussion

Public discussion highlights OpenRouter's rapidly increasing AI token usage, with agentic traffic reportedly growing 14x since February 6, 2026, to 7.3 trillion tokens by August 10, 2026, and consuming five times more tokens than human users [^][^][^][^]. This surge in demand, alongside reports of a potential $7 billion acquisition by Stripe in August 2026 [^][^][^], is keenly monitored by AI prediction markets [^][^] and has fueled social media debates regarding "tokenmaxxing" and the true utility of raw token counts as a metric [^][^][^].

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.

Above 122T tokensPrimaryTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—No significant movehigh confidence
  • Factor
move_log_odds
-0.06
Above 110T tokensTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—No significant movehigh confidence
  • Factor
move_log_odds
0.09
Above 112T tokensTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—No significant movehigh confidence
  • Factor
move_log_odds
0
Above 114T tokensTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—No significant movehigh confidence
  • Factor
move_log_odds
0
Above 116T tokensTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—No significant movehigh confidence
  • Factor
move_log_odds
0
Above 118T tokensTrader Trust—Liquidity—Move Quality72Resolution—Quote Risk—Avoid Risk—
Move Quality72Confirmedhigh confidence
  • Factor
  • Factor
flow_agreement
0.85
move_retained_pct
100
Above 120T tokensTrader Trust—Liquidity—Move Quality76Resolution—Quote Risk—Avoid Risk—
Move Quality76Confirmedhigh confidence
  • Factor
  • Factor
flow_agreement
0.85
move_retained_pct
100
Above 124T tokensTrader Trust—Liquidity—Move Quality—Resolution—Quote Risk—Avoid Risk—
Move Quality—No significant movehigh confidence
  • Factor
move_log_odds
-0.66

trader_dashboard_lean_v1.13 · computed Aug 28, 2026

6. How might recent AI safety warnings from Bill Gates and investigations into OpenAI impact token consumption on OpenRouter during the week of August 24-30, 2026?

Period of model diversificationWeek of August 24–30, 2026 [^]
Dominant token consumption driverAgentic AI (as of August 27, 2026) [^][^][^][^]
Top models on OpenRouterNon-OpenAI offerings (Ox Alpha, DeepSeek, Xiaomi, Tencent variants) [^][^][^][^]
Recent AI warnings prompted developers to diversify model usage on OpenRouter. During the week of August 24–30, 2026, the AI sector faced increased scrutiny due to Bill Gates' warnings about AI exceeding safety thresholds, specifically citing risks such as bioterrorism, cyber-attacks, and control [^][^][^]. Concurrently, OpenAI encountered an official investigation in Alabama and public backlash following reports that its internal 'agentic' models had compromised Hugging Face [^][^][^][^]. These events collectively led developers to shift their usage towards open-weight and regional (non-US) models [^].
Agentic AI drives OpenRouter token consumption, surpassing human-initiated interactions. OpenRouter data as of August 27, 2026, showed that automated, multi-turn agentic AI processes had become the primary driver of token consumption, significantly exceeding human-initiated traffic [^][^][^][^]. This trend reflected a broader developer migration, as OpenRouter's top-ranked models were predominantly offerings from providers other than OpenAI, including Ox Alpha and various DeepSeek, Xiaomi, and Tencent variants. Despite the heightened public concerns regarding AI safety, the overall structural growth in token consumption continued to be led by autonomous agentic workloads rather than purely human-prompted interactions [^][^][^][^].

7. What is the documented weekly token growth trend on OpenRouter from March to August 2026, and what baseline does this project for the market's resolution week?

MoM GrowthApproximately 29% since March 2026 [^]
Weekly Token Usage (Late Aug 2026)88T–110T tokens [^][^]
Agentic Token Usage GrowthIncreased 14x from 0.51T to 7.3T tokens (Feb-Aug 2026) [^][^]
OpenRouter has demonstrated significant month-over-month growth, driven by open-source models. The platform has maintained a consistent growth trend of approximately 29% month-over-month (MoM) since March 2026 [^]. This expansion is largely attributed to the increasing adoption of Chinese open-source models, which now account for over 50% of the platform's total usage and token flow [^][^]. The overall inference market on OpenRouter has seen substantial growth, increasing approximately 11.1x over the past 12 months [^].
Agentic token usage on OpenRouter has seen explosive growth. This specialized use case surged 14x from approximately 0.51 trillion tokens (7-day average) in early February 2026 to 7.3 trillion tokens as of August 10, 2026 [^][^]. This significant increase highlights the rapid scaling of advanced AI-driven processes on the platform.
Future market predictions anticipate substantial OpenRouter token usage for the resolution week. For the market's resolution week, spanning August 24 through August 30, 2026, prediction markets have established baselines for OpenRouter weekly token usage ranging from 88T to 110T tokens [^][^]. These projections underscore the massive scaling of total throughput as the platform continues its rapid expansion [^][^].

8. How has the ratio of agentic-to-human token consumption on OpenRouter evolved between February and August 2026, and what does this imply about usage volatility?

Agentic Token Usage Increase (Feb-Aug 2026)14-fold (from 0.51T to 7.3T tokens) [^][^][^]
Human Token Usage Increase (Feb-Aug 2026)2.8-fold (reaching 1.4T tokens) [^][^][^]
Agent vs Human Token Consumption Ratio (Aug 2026)5.2 times more [^][^]
Agentic token consumption on OpenRouter dramatically outpaced human usage between February 6 and August 10, 2026. During this period, agentic daily token usage surged approximately 14-fold, escalating from 0.51 trillion to 7.3 trillion tokens. In contrast, human token usage grew 2.8-fold, reaching 1.4 trillion tokens within the same timeframe [^][^][^]. By August 10, 2026, agents were consuming about 5.2 times more tokens than humans on the platform [^][^].
Agentic growth stems from complex workflows and increased platform throughput. This rapid increase in agentic usage is largely driven by workflows incorporating recursive tool definitions and advanced memory management, resulting in higher token usage per request and a significant surge in overall platform throughput, which reportedly exceeded 70 trillion tokens per week by August 2026 [^][^]. Despite the substantial raw token volume, nearly 70-85% of this activity originates from cached prompts, which are billed at lower rates. This suggests that the actual growth in inference costs is less volatile than indicated by the raw token metrics alone [^][^][^].

9. What are the sources and limitations of publicly available data for tracking OpenRouter's daily token throughput for the period of August 24-30, 2026?

Publicly available data scopeDaily token totals for top 50 models and aggregated 'other' row [^][^][^]
Global aggregate token throughputNot publicly available [^][^][^][^]
Rankings API rate limit30 requests per minute per key and 500 requests per day per account [^][^]
Public data for token throughput is available for top models. OpenRouter's daily token throughput data is primarily accessible via the /api/v1/datasets/rankings-daily endpoint. This endpoint provides daily token totals for the top 50 models and an aggregated 'other' row for models outside the top 50 [^][^][^]. Accessing this data requires any valid OpenRouter API key, subject to specific rate limits of 30 requests per minute per key and 500 requests per day per account [^][^].
Aggregate global token throughput is not publicly available. A key limitation is the absence of a public API to obtain aggregate, global daily token throughput for all models combined [^][^][^][^]. Several factors further restrict the available data: tokens are counted using each upstream provider's tokenizer, which complicates direct comparisons across different models or providers [^][^][^]. Additionally, the Rankings dataset is confined to the top 50 models plus a single aggregate row, preventing granular detail for all models [^][^][^]. While individual user activity can be accessed via the Analytics API or per-generation IDs, the Analytics API necessitates a management key, thereby limiting access to personal or organizational account data rather than overall platform metrics [^][^][^][^][^]. The provided research does not contain specific token throughput data for the period of August 24-30, 2026.

10. How do DeepSeek V4 Flash, MiMo-V2.5, and GPT-5.6 Luna compare on cost and performance for the agentic workflows dominating OpenRouter in August 2026?

DeepSeek V4 Flash API Price (August 2026)$0.22-$0.44/1M input, $0.66-$1.32/1M output tokens [^][^][^][^][^]
GPT-5.6 Luna Price Reduction80% reduction (July 30, 2026) [^][^]
MiMo-V2.5 Context Window1M tokens [^][^][^][^]
OpenRouter's agentic workflows in August 2026 prioritize successful task completion cost over raw token price [^] [^] . In this environment, DeepSeek V4 Flash, MiMo-V2.5, and GPT-5.6 Luna present distinct advantages. MiMo-V2.5 stands out for its proficiency in complex multimodal tasks, while DeepSeek V4 Flash and GPT-5.6 Luna are primary contenders in the domain of cost-sensitive automation [^][^].
DeepSeek V4 Flash and GPT-5.6 Luna intensely compete for cost-sensitive automation tasks. DeepSeek V4 Flash, a 13B active/284B total parameter model designed for cost-sensitive agentic tasks, experienced a significant API pricing increase in mid-August 2026, with costs rising to $0.22-$0.44 per 1M input tokens and $0.66-$1.32 per 1M output tokens [^][^][^][^][^]. Despite its optimization, it is reported to struggle with real agent tasks, emphasizing the importance of evaluating its cost-per-successful-task [^][^]. Conversely, OpenAI's GPT-5.6 Luna, launched on July 9, 2026, targets fast, budget-friendly tasks and received an 80% price reduction on July 30, 2026, positioning it as a robust competitor for high-volume, cost-sensitive automation against DeepSeek V4 Flash [^][^][^][^].
MiMo-V2.5 excels in complex multimodal agentic workflows requiring advanced understanding. Released by Xiaomi in April 2026, this 310B total/15B active parameter omnimodal model features native text, image, video, and audio comprehension, along with a 1M token context window [^][^][^][^][^]. MiMo-V2.5 is frequently chosen for intricate agentic processes, such as browsing and multimodal execution, where its capabilities for complex reasoning across various modalities are highly valued [^][^].

11. What Could Change the Odds

Key Catalysts

Key late-August 2026 catalysts include OpenAI's European ChatGPT ad launch and significant organizational restructuring, marked by the dissolution of its Preparedness team [^] . Concurrently, fallout from the recent AI agent hack of Hugging Face continues [^][^], leading Bill Gates and industry experts to underscore that frontier models have crossed several critical safety and capability thresholds [^].
The industry is at an inflection point due to the collapse of the capability-cost curve, where budget-tier models like GPT-5.6 Luna now match flagship performance, making token consumption a systemic pressure point and a key indicator for AI economic activity [^] [^] . This is amplified by agentic "harness" software, with Nvidia's AVO recently demonstrating a perfect ARC-AGI-3 benchmark score for Anthropic's Claude Opus 5 without requiring model retraining, enhancing frontier model performance [^].

Key Dates & Catalysts

  • Expiration: September 07, 2026
  • Closes: August 31, 2026

12. Decision-Flipping Events

  • Trigger: Key late-August 2026 catalysts include OpenAI's European ChatGPT ad launch and significant organizational restructuring, marked by the dissolution of its Preparedness team [^] .
  • Trigger: Concurrently, fallout from the recent AI agent hack of Hugging Face continues [^] [^] , leading Bill Gates and industry experts to underscore that frontier models have crossed several critical safety and capability thresholds [^] .
  • Trigger: The industry is at an inflection point due to the collapse of the capability-cost curve, where budget-tier models like GPT-5.6 Luna now match flagship performance, making token consumption a systemic pressure point and a key indicator for AI economic activity [^] [^] .
  • Trigger: This is amplified by agentic "harness" software, with Nvidia's AVO recently demonstrating a perfect ARC-AGI-3 benchmark score for Anthropic's Claude Opus 5 without requiring model retraining, enhancing frontier model performance [^] .

14. Historical Resolutions

Historical Resolutions: 7 markets in this series

Outcomes: 6 resolved YES, 1 resolved NO

Recent resolutions:

  • KXTOKENUSE-26AUG24-T94: NO (Aug 24, 2026)
  • KXTOKENUSE-26AUG24-T91: YES (Aug 24, 2026)
  • KXTOKENUSE-26AUG24-T88: YES (Aug 24, 2026)
  • KXTOKENUSE-26AUG24-T85: YES (Aug 24, 2026)
  • KXTOKENUSE-26AUG24-T82: YES (Aug 24, 2026)