openrouter lab share Odds: DeepSeek 99%, OpenAI 97% for #2
The openrouter lab share markets look settled: Polymarket priced DeepSeek near 99.4% for #1 and OpenAI near 97.3% for #2 this week at 06:02 UTC.
Early story. Some claims here are not officially confirmed yet. We update this post as it confirms.

- Polymarket #1 lab: DeepSeek ~99.4% Yes, MiniMax ~1.7%; #2 lab: OpenAI ~97.3%, Google ~3.4%, MiniMax ~1.1%.
- OpenRouter top models by tokens: DeepSeek V4.1 Flash 1017 at about 42.4T tokens (+96%), DeepSeek V4.1 Flash at 15.9T, GLM 5.3 Flash at 15.6T, GPT-6 Luna at 11.4T.
- The catch: these markets resolve on OpenRouter's Market Share chart, which counts requests, not tokens.
in this block
The openrouter lab share markets for this week look almost settled, and the interesting part is why. At 06:02 UTC (09:02 MSK) on October 30, Polymarket priced DeepSeek at about 99.4% to finish first in OpenRouter market share for the week of October 26 to November 1, and OpenAI at about 97.3% to finish second. OpenRouter's own public leaderboard, with usage data through October 29, shows DeepSeek's V4.1 Flash 1017 far ahead on tokens processed.
What actually happened
OpenRouter routes traffic from developers to hundreds of models and publishes the stats. Its rankings page says the Top Models chart and LLM Leaderboard rank models by tokens processed, counting prompt and completion tokens, while the Market Share chart counts requests. Neither measures users, spend, or quality.
On the token board, DeepSeek dominates. V4.1 Flash 1017 sits first with about 42.4 trillion tokens and almost doubled versus the prior period. The original V4.1 Flash is second at about 15.9 trillion. Z.ai's GLM 5.3 Flash is third, OpenAI's GPT-6 Luna fourth at about 11.4 trillion (+86%), and Tencent's Hy4 preview fifth. MiniMax M3.1 Flash shows growth above 999% from a small base, at about 4.1 trillion.
The rest of the top 10 is a who's who of cheap, fast models. Anthropic's Claude Opus 5.5 sits sixth at about 7.8 trillion tokens, StepFun's Step 5 Preview seventh with growth around 262%, and an older DeepSeek V4 Flash build rounds out the list. OpenRouter publishes this data under a CC BY 4.0 license, which is why so many dashboards and prediction markets can lean on it.
What Polymarket thinks
Two markets matter: "#1 AI lab by OpenRouter market share week of October 26?" and the #2 version. Both resolve on the percentage shown for each lab in OpenRouter's Market Share chart for October 26 to November 1, once the next week's figures first appear. "Stealth" and "others" groupings are skipped.
At our 06:02 UTC snapshot:
- 1: DeepSeek about 99.4% Yes (last trade around 95.6¢), MiniMax about 1.7%, everyone else under 0.5%.
- 2: OpenAI about 97.3% Yes, Google about 3.4%, MiniMax about 1.1%, Z.ai about 0.5%.
Volumes are small, about $12.7K on the #1 event and $19.1K on the #2 event, so a single trader can still nudge prices.
Tokens tell you who is busy. Requests decide who gets paid on this market.
The tokens versus requests trap
Here is the nuance most people miss. If you add the two Z.ai GLM models in OpenRouter's top 10 by tokens, they exceed GPT-6 Luna's token total. Yet traders price OpenAI near 97% for second place. That is not a contradiction: the openrouter lab share contract uses requests, and a model that handles many short calls can rank differently from one that chews huge contexts.
So the token chart is a hint, not the answer. DeepSeek's lead looks wide enough that both views agree on first place. Second place is where the measurement choice really matters, and the market is trusting the request chart traders can see.
We covered the model behind DeepSeek's surge in deepseek v4.1 flash 1017, and MiniMax's fast climber in minimax m3.1 flash. Both show up in this week's openrouter lab share story.
What to do as a reader
If you want to trade these, open OpenRouter's Market Share chart, switch it to Percentage, and look at the current week directly. Do not rely on the token leaderboard alone.
Remember the timing rule: resolution waits until the November 2 week first appears. Data can shift in the last days of a week, especially if a new model launches and gets routed heavily.
At 97% to 99%, the upside is tiny and the downside is your whole stake. These are carry trades, not lottery tickets, and thin books make exits uncertain.
Nothing here is investment advice. Prediction markets can resolve against you even when the odds look certain.
Sources: OpenRouter rankings, Polymarket #1 lab, Polymarket #2 lab.
Not financial advice. DYOR, ser.