Microsoft announced MAI-Thinking-1, its first in-house reasoning model, at Build 2026 in early June as part of a seven-model MAI family trained from scratch on clean enterprise data without distillation. The 35B-active-parameter MoE model (roughly 1T total) with 256K context targets multi-step reasoning, math, and coding, posting 97% on AIME 2025 and competitive SWE-Bench Pro scores while earning preference over Claude Sonnet 4.6 in blind evaluations. It moved to public preview in Microsoft Foundry by mid-August, underscoring rapid iteration and Microsoft's push for Azure-native independence from external frontier labs. Traders are monitoring integration timelines into Copilot and developer platforms, plus any follow-on "hill-climbing" releases, as the pace of internal progress and benchmark updates could accelerate a successor.
Ringkasan eksperimental yang dihasilkan AI dengan referensi data Polymarket. Ini bukan saran trading dan tidak berperan dalam bagaimana pasar ini diselesaikan. · DiperbaruiSeptember 30
27%
December 31
49%
$489 Vol.
September 30
27%
December 31
49%
A “successor to MAI-Thinking-1” refers to any model explicitly designated by Microsoft as part of its MAI (Microsoft AI) model family and as a thinking or reasoning model, or presented as Microsoft’s next general-purpose reasoning model or widely regarded as the successor to MAI-Thinking-1.
Task-specialized models, cost-efficiency variants, or similar releases will not count unless they are explicitly designated as thinking or reasoning models or widely regarded as successors to MAI-Thinking-1.
Products outside of the MAI family or not designated as thinking or reasoning models will NOT qualify.
For this market to resolve to "Yes," a qualifying model must be launched and publicly accessible, including via open beta or open rolling waitlist signups. A closed beta or any form of private access will not suffice.
The release must be either clearly defined and publicly announced by Microsoft as being accessible to the general public or otherwise made publicly accessible and explicitly labeled within the company’s official website. Labeling errors, placeholder text, or version names displayed on the website that do not correspond to a model that is actually accessible to the general public will not qualify.
The primary resolution source for this market will be official information from Microsoft (e.g., https://microsoft.ai/models), with additional verification from a consensus of credible reporting.
Pasar Dibuka: Jul 11, 2026, 2:45 PM ET
Resolver
0x65070BE91...A “successor to MAI-Thinking-1” refers to any model explicitly designated by Microsoft as part of its MAI (Microsoft AI) model family and as a thinking or reasoning model, or presented as Microsoft’s next general-purpose reasoning model or widely regarded as the successor to MAI-Thinking-1.
Task-specialized models, cost-efficiency variants, or similar releases will not count unless they are explicitly designated as thinking or reasoning models or widely regarded as successors to MAI-Thinking-1.
Products outside of the MAI family or not designated as thinking or reasoning models will NOT qualify.
For this market to resolve to "Yes," a qualifying model must be launched and publicly accessible, including via open beta or open rolling waitlist signups. A closed beta or any form of private access will not suffice.
The release must be either clearly defined and publicly announced by Microsoft as being accessible to the general public or otherwise made publicly accessible and explicitly labeled within the company’s official website. Labeling errors, placeholder text, or version names displayed on the website that do not correspond to a model that is actually accessible to the general public will not qualify.
The primary resolution source for this market will be official information from Microsoft (e.g., https://microsoft.ai/models), with additional verification from a consensus of credible reporting.
Resolver
0x65070BE91...Microsoft announced MAI-Thinking-1, its first in-house reasoning model, at Build 2026 in early June as part of a seven-model MAI family trained from scratch on clean enterprise data without distillation. The 35B-active-parameter MoE model (roughly 1T total) with 256K context targets multi-step reasoning, math, and coding, posting 97% on AIME 2025 and competitive SWE-Bench Pro scores while earning preference over Claude Sonnet 4.6 in blind evaluations. It moved to public preview in Microsoft Foundry by mid-August, underscoring rapid iteration and Microsoft's push for Azure-native independence from external frontier labs. Traders are monitoring integration timelines into Copilot and developer platforms, plus any follow-on "hill-climbing" releases, as the pace of internal progress and benchmark updates could accelerate a successor.
Ringkasan eksperimental yang dihasilkan AI dengan referensi data Polymarket. Ini bukan saran trading dan tidak berperan dalam bagaimana pasar ini diselesaikan. · Diperbarui



Hati-hati dengan link eksternal.
Hati-hati dengan link eksternal.
Pertanyaan yang Sering Diajukan