Nex-N2.5-Pro

3 sources, 0 independent of the vendor: centre 32; the sources disagree widely, so shown as a band.

• 31.9 SWE-bench Pro (public): 61.2 · vendor · read against Kimi K3, GLM-5.3, DeepSeek V4 Pro 0813, Qwen3.8 Max, GPT-5.6 Sol in the same table
• 32 Terminal-Bench 2.1: 82.7 · vendor · read against Kimi K3, GLM-5.3, DeepSeek V4 Pro 0813, Qwen3.8 Max, Claude Opus 5, GPT-5.6 Sol in the same table
• 21.9 toolathlon verified: 68.5 · vendor
• browsecomp: 89.7 · vendor · not used: no reference models for this benchmark
• deepswe v1 1: 55.8 · vendor · not used: no reference runs on the same harness
• osworld verified: 82.2 · vendor · not used: no reference models for this benchmark

how FTI is computed