AI model library — official specs, cutoffs & comparisons

31 matching models

MicrosoftAvailable

MAI-Code-1.1-Flash

MAI

Open weights

Context: 256,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

MAI-Code-1-Flash

MAI · MAI-Code-1-Flash

GitHub Copilot

Context: Not documented
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-reasoning-vision-15B

Phi

Open weights

Context: 16,384 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftPreview

MAI-1-preview

MAI

API

Context: Not documented
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-mini-flash-reasoning

Phi

Open weights

Context: 64,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-reasoning

Phi

Open weights

Context: 32,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-reasoning-plus

Phi

Open weights

Context: 32,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-mini-reasoning

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-mini-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-4-multimodal-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

phi-4

Phi

Open weights

Context: 16,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3.5-mini-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3.5-MoE-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3-medium-128k-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3-medium-4k-instruct

Phi

Open weights

Context: 4,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3-small-128k-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3-small-8k-instruct

Phi

Open weights

Context: 8,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

Phi-3-vision-128k-instruct

Phi

Open weights

Context: 128,000 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

MAI-Cyber-1-Flash

MAI · MAI-Cyber-1-Flash

MDASH private preview for approved defenders

Context: 256k tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

MAI-DS-R1

MAI

Open weights

Context: Not documented
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftPreview

MAI-Thinking-1

MAI · MAI-Thinking-1

API

Context: 256K tokens
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

phi-1

Phi

Open weights

Context: Not documented
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

phi-1_5

Phi

Open weights

Context: Not documented
Max. output: Not documented
Knowledge cutoff: Not documented

MicrosoftListed by publisher

phi-2

Phi

Open weights

Context: 2,048 tokens
Max. output: Not documented
Knowledge cutoff: Not documented

Go beyond the spec sheet

The decisions behind the numbers.

Source-checked guides to pricing, usable context, and access restrictions.

Claude Sonnet 4.6 pricing: API costs, caching and real workload math GPT-5.4 context window: the 1.05M limit, output budget and cost threshold Gemini 2.5 Pro context window: input limits, access and long-document costs GPT-5.4 mini context window: 400K capacity and practical task budgets