Language modelPreview

MAI-Thinking-1

Microsoft AI

Released: August 12, 2026
Data date: October 3, 2026

MAI-Thinking-1 was trained without distillation from other providers’ models, according to Microsoft. The company uses its own traceable training data and executable coding environments where real tests grade outcomes. The MoE model activates 35 billion of roughly 1 trillion parameters.

It has been available in public preview on Microsoft Foundry since August 12, 2026. Microsoft specifies a 256K-token context window.

Specifications and access

Specification	Value and source
Release	August 12, 2026Source
Access	Public preview in Microsoft FoundrySource
Focus	Reasoning for mathematics, knowledge, and codingSource
Parameters	About 1 trillion total according to Microsoft, 35 billion active (mixture of experts)Source
Context window	256K tokens according to MicrosoftSource
Capabilities	Function calling, developer instructions, and a Chat Completions-compatible APISource

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

AIME 2025 (Show measurement, test conditions, and source)

Source value: 97
Score: 97
Metric: AIME 2025
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

AIME 2026 (Show measurement, test conditions, and source)

Source value: 94.5
Score: 94.5
Metric: AIME 2026
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

HMMT Feb 2026 (Show measurement, test conditions, and source)

Source value: 84.9
Score: 84.9
Metric: HMMT Feb 2026
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

GPQA Diamond (Show measurement, test conditions, and source)

Source value: 84.2
Score: 84.2
Metric: GPQA Diamond
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

LCB v6 (Show measurement, test conditions, and source)

Source value: 87.7
Score: 87.7
Metric: LCB v6
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

Terminal Bench 2.0 (Show measurement, test conditions, and source)

Source value: 46
Score: 46
Metric: Terminal Bench 2.0
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

SWE-Bench Verified (Show measurement, test conditions, and source)

Source value: 73.5
Score: 73.5
Metric: SWE-Bench Verified
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

SWE-Bench Pro (Show measurement, test conditions, and source)

Source value: 52.8
Score: 52.8
Metric: SWE-Bench Pro
Unit: %
Category: source-specific
Direction: Higher is better
Source type: vendor-reported
Evaluator: Microsoft announcement Table 1 pixels
Status: active
Retrieved at: 2026-10-04
Methodology: Vendor-reported result. Post-trained model evaluation; percentages. Agentic coding uses 256K total context; others max output 256K. Separate blind human preference study has 1,276 tasks.
Context: Source-specific observation; it is not a shared comparison cohort. Microsoft self-reported table; peer numbers taken from respective official cards

Source Microsoft announcement Table 1 pixels

Sources and data date

Every statement links to its underlying documentation or leaderboard.

Type	Evidence and data date
Research status	Research date October 4, 2026. 1 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Additional source	Microsoft AI MAI-Thinking-1 announcement (retrieved October 3, 2026; October 4, 2026) · Microsoft AI MAI-Thinking-1 announcement · Editorial description reviewed October 4, 2026