MAI-Thinking-1

Coding

• SWE-bench Pro#23 of 31
  52.8%vendor256k context; bash + str_replace toolsMAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: Claude Opus 5.5 at 89.9%

• SWE-bench Verified#8 of 16
  73.5%vendor256k contextMAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: Claude Sonnet 5 at 85.2%

• Terminal-Bench 2.x#63 of 103
  46%vendorTerminal-Bench 2.0; bash tool onlyMAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: Claude Fable 5.1 at 91.4%

• LiveCodeBench#6 of 56
  87.7%vendorv6MAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: Fugu at 92.9%

Reasoning and knowledge

• AIME 2026#6 of 13
  94.5%vendorAIME 2026MAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: ERNIE 5.1 at 99.6%

• HMMT February 2026#5 of 8
  84.9%vendorHMMT Feb 2026MAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: GLM-5.2 at 92.5%

• GPQA Diamond#62 of 142
  84.2%vendorMAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: GPT-6 Astra at 96.1%

• MMLU-Pro#3 of 14
  85%vendorMAI-Thinking-1 technical report (Table 11/12) ↗

  Leader: Nemotron 3 Ultra 550B A55B at 86.8%

Other published results