ModelsCompareBest forBenchmarksStatusPricingAPI

MAI-Thinking-1

Microsoft·Pending·Released 2026-08-12text
Context256k
Max output64k
Input / 1M-
Output / 1M-
Cached in-
Benchmarks15

Benchmarks

Cited public sources · rank is exact-benchmark, apples-to-apples

AdvancedIF85%
AIME 2025#2/1597%
AIME 2026#2/294.5%
BFCL v372%
GPQA Diamond#38/7084.2%
GraphWalks (<=128K)90%
HMMT February 202684.9%
IFBench#4/569%
LiveCodeBench v6#3/1187.7%
MMLU-Pro#8/2985%
MultiChallenge#3/353%
Terminal-Bench 2.0#12/1346%

Access MAI-Thinking-1 and every other model through one endpoint with automatic failover: Respan gateway.