SYLIN

DeepSWE

Benchmark · Text & code

MiMo 2.6 Pro near GPT-6 Astra and above Fable 5 on this coding benchmark

Evidence

  1. Sep 24
    V4.1 Flash scores 74.2, topping other open models
  2. Jul 27
    Described as a leading measure of agentic software engineering; Anthropic's chart grayed out GPT-5.6, which scored highest in its table
  3. Jul 16
    Fable 5 and GPT Sol score ahead of Kimi K3 on DeepSWE

Alternatives

HyClaimed to lead Terminal-Bench 2.1 and DeepSWE-style agentic tests, ahead of GLM 5.3, Kimi K3 and DeepSeek V4 Pro