Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Microsoft AI / MAI

MAI-Thinking-1

Reasoning and coding in Microsoft Foundry.

Language models (LLMs)
previewClosed weightsReasoningCoding
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
13 of 18 specification fields collected
Context window256K256,000 tokens
Maximum output64K64,000 tokens
Input / 1M tokens$2.00Microsoft Foundry
Output / 1M tokens$8.00USD · observed 14 Oct 2026

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
256,000 tokensSource for Context / input limit (opens in a new tab)
Maximum output
64,000 tokensSource for Maximum output (opens in a new tab)

Includes reasoning tokens and must fit within the remaining context budget.

Input modalities
TextSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)
Reasoning
SupportedSource for Reasoning (opens in a new tab)

Optional opaque encrypted reasoning state can be carried across turns; it is not a readable reasoning trace.

Tool calling
SupportedSource for Tool calling (opens in a new tab)

Function tools only.

Structured output
Not collected
Streaming
SupportedSource for Streaming (opens in a new tab)

Server-sent chat completion chunks.

Batch processing
Not collected
Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
2026-08-12Source for Release date (opens in a new tab)
Parameters
35B active; approximately 1T totalSource for Parameters (opens in a new tab)
Architecture
Mixture of expertsSource for Architecture (opens in a new tab)
Languages
Not collected
Context caching
SupportedSource for Context caching (opens in a new tab)

Documented prefix-cache token accounting and a published cached-input rate; retention and minimum prefix size remain unknown.

Model revision
2026-06-01Source for Model revision (opens in a new tab)

Foundry model version, not launch date.

Request quotas
Global Standard: default low tier 0 TPM / 0 RPM; medium 100,000 TPM / 100 RPM; high 250,000 TPM / 250 RPM.Source for Request quotas (opens in a new tab)

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Microsoft Foundry

Provider source (opens in a new tab)
API model ID
Not collected

Use your Microsoft Foundry deployment name in model; MAI-Thinking-1 is the catalog model name.

Access
Public preview, Global Standard. POST /mai/v1/chat/completions. Authenticate with an API key or Microsoft Entra ID. Use max_completion_tokens, not max_tokens.
Regions
Global Standard deployment; exact resource-region availability not collected.
Context caching
SupportedSource for Context caching (opens in a new tab)

Documented prefix-cache token accounting and a published cached-input rate; retention and minimum prefix size remain unknown.

Reasoning
SupportedSource for Reasoning (opens in a new tab)

Optional opaque encrypted reasoning state can be carried across turns; it is not a readable reasoning trace.

Tool calling
SupportedSource for Tool calling (opens in a new tab)

Function tools only.

Streaming
SupportedSource for Streaming (opens in a new tab)

Server-sent chat completion chunks.

INPUT / 1M TOKENS$2.00
OUTPUT / 1M TOKENS$8.00

Global Standard USD list-price snapshot for the us-east pricing selection. Region, agreement, taxes and additional services may alter the bill. Generated reasoning counts toward completion tokens. Cached input price: $0.20 / 1M tokens (base tier).

USD · observed 14 Oct 2026 · Pricing evidence (opens in a new tab)

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

Microsoft Foundry

Documented
Support evidence (opens in a new tab)
How it works
Prefix cachingSource for How it works (opens in a new tab)
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
usage.prompt_tokens_details.cached_tokens counts cached prompt tokens.Source for Usage fields (opens in a new tab)
Billing rules
USD 0.20 per 1M cached input tokens for the us-east pricing selection; ordinary input USD 2 and output USD 8 per 1M tokens.Source for Billing rules (opens in a new tab)
Limits & data boundaries
Not collected
Cache reads, writes & storage
ChargeApplies toRate / unit
Cache read1Text tokens · Base tier$0.20 / 1M tokensPricing evidence (opens in a new tab)

Rate conditions

  1. Global Standard USD list-price snapshot for the us-east pricing selection. Region, agreement, taxes and additional services may alter the bill. Generated reasoning counts toward completion tokens.

Observed 14 Oct 2026. Write rates and storage fees are different charges; use the billing rules above.

Caching reviewed 14 Oct 2026

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Proprietary hosted release documented by its developer; no public weights identified in the reviewed release documentation.

Classification evidence (opens in a new tab)

Sources reviewed