Microsoft model product
MAI Thinking 1
MAI-Thinking-1 is Microsoft AI's first in-house reasoning model, a 35B-active / ~1T-total parameter sparse Mixture of Experts model (base model MAI-Base-1) trained from scratch without distillation from third-party models.
Updated Aug 17, 2026. Default version: MAI-Thinking-1
LLMBoard Score54.5MAI-Thinking-1
Coverage100%17 benchmark families
Context windowN/ATokens
Official input priceN/AOfficial price unavailable
MAI Thinking 1 Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
MAI Thinking 1 Benchmark Results
Benchmark scores for MAI-Thinking-1.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkAdvancedIF | Score85.0% | Rank01 | Participants2 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkAIR-Bench | Score88.0% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkCorpusQA | Score82.0% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkCyberSecEval 4 | Score63.0% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkGraphWalks | Score90.0% | Rank01 | Participants3 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkLongFact | Score98.0% | Rank01 | Participants1 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkSimpleQA Verified | Score31.0% | Rank01 | Participants2 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkTruthfulQA | Score88.0% | Rank01 | Participants18 | Percentile100.0% | EvidenceC | Evaluated |
| BenchmarkBFCL-v3 | Score72.0% | Rank04 | Participants19 | Percentile83.3% | EvidenceC | Evaluated |
| BenchmarkLongBench v2 | Score61.0% | Rank07 | Participants17 | Percentile62.5% | EvidenceC | Evaluated |
| BenchmarkLiveCodeBench v6 | Score87.7% | Rank08 | Participants56 | Percentile87.3% | EvidenceC | Evaluated |
| BenchmarkAIME 2026 | Score94.5% | Rank09 | Participants21 | Percentile60.0% | EvidenceC | Evaluated |
| BenchmarkHealthBench Professional | Score35.0% | Rank09 | Participants9 | Percentile0.0% | EvidenceC | Evaluated |
| BenchmarkHMMT Feb 26 | Score84.9% | Rank09 | Participants12 | Percentile27.3% | EvidenceC | Evaluated |
| BenchmarkMedXpertQA | Score43.0% | Rank09 | Participants12 | Percentile27.3% | EvidenceC | Evaluated |
| BenchmarkAIME 2025 | Score97.0% | Rank15 | Participants115 | Percentile87.7% | EvidenceC | Evaluated |
| BenchmarkMulti-Challenge | Score53.0% | Rank16 | Participants29 | Percentile46.4% | EvidenceC | Evaluated |
| BenchmarkIFBench | Score69.0% | Rank23 | Participants34 | Percentile33.3% | EvidenceC | Evaluated |
| BenchmarkMMLU-Pro | Score85.0% | Rank24 | Participants134 | Percentile82.7% | EvidenceC | Evaluated |
| BenchmarkSWE-Bench Pro | Score52.8% | Rank40 | Participants50 | Percentile20.4% | EvidenceC | Evaluated |
| BenchmarkTerminal-Bench 2.0 | Score46.0% | Rank43 | Participants51 | Percentile16.0% | EvidenceC | Evaluated |
| BenchmarkSWE-Bench Verified | Score73.5% | Rank44 | Participants111 | Percentile60.9% | EvidenceC | Evaluated |
| BenchmarkGPQA | Score84.2% | Rank64 | Participants239 | Percentile73.5% | EvidenceC | Evaluated |
MAI Thinking 1 Arena Results
Preference and agent-evaluation results for the default version.
No Arena results
The default version does not have a matching Arena result yet.
MAI Thinking 1 Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- N/A
- Tracked offerings
- 0
No provider prices
The default version has no current input or output token prices.
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
MAI Thinking 1 Runtime Performance
Provider-specific output speed and catalog latency for MAI-Thinking-1. Runtime does not affect the capability score.
No runtime data
Browse runtime rankingsNo provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
MAI Thinking 1 Specifications
Technical details for the model's default version.
- Version
- MAI-Thinking-1
- Released
- Jun 2, 2026
- Knowledge cutoff
- Unknown
- Parameters
- 1T
- Context window
- N/A
- Max output
- N/A
- Inputs
- text
- Outputs
- text
- Open weights
- No
- License
- Proprietary
MAI Thinking 1 Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionMAI-Thinking-1 | Released | LLMBoard54.5 | Parameters1T | ContextN/A | Max outputN/A | Open weightsNo | LicenseProprietary |
MAI Thinking 1 vs nearby models
Open a comparison with the three ranked models immediately above and below this model.
Models similar to MAI Thinking 1
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
MI
MAI Code 1 Flash
Microsoft
MI
MAI Code 1.1 Flash
Microsoft
MI
Phi 4 Reasoning Plus
Microsoft
MI
Phi 4 Reasoning
Microsoft
MI
Phi 4 Mini Reasoning
Microsoft
MI
Phi 4
Microsoft
What is MAI Thinking 1?
Key information about MAI Thinking 1 and its available data.
MAI-Thinking-1 is Microsoft AI's first in-house reasoning model, a 35B-active / ~1T-total parameter sparse Mixture of Experts model (base model MAI-Base-1) trained from scratch without distillation from third-party models.
55T mid-training tokens), then post-trained via reinforcement learning across STEM, agentic coding, and helpfulness/safety specialists consolidated into a single model. 0% on AIME 2025. 6 in blind human side-by-side evaluations.
Data as of 2026-08-17.
FAQ
Common questions about MAI Thinking 1.
When was MAI Thinking 1 released?
MAI Thinking 1's default version was released on Jun 2, 2026.
How much does MAI Thinking 1 cost?
No official standard PAYG price is currently available for MAI Thinking 1.
Who created MAI Thinking 1?
MAI Thinking 1 was created by Microsoft.
What is the context window for MAI Thinking 1?
A context window is not available for the default version.
Is MAI Thinking 1 open weight?
No. The default version is not marked as having publicly available weights.
How many API providers offer MAI Thinking 1?
No provider offering is currently linked to the default version.
What models should I compare MAI Thinking 1 with?
Nearby ranked alternatives include MiniMax M2.5, Qwen3.6 35B A3B, Grok 3.